Self-renewal and pluripotency are defining properties of embryonic stem cells (ESCs). They refer, respectively, to the ability to proliferate indefinitely without commitment in vitro and to the capacity to differentiate into cell lineages belonging to the three embryonic germ layers1,2,3. ESCs are derived from the inner cell mass (ICM) of pre-implantation embryos1,2, but alternative approaches — such as nuclear transfer, cell fusion or direct reprogramming (reviewed in Ref. 4) — are now available that allow the generation of pluripotent stem-cell lines directly from differentiated adult somatic tissue. These have widened the range of applications in which these cells can be used5,6,7. Among these methodologies, direct reprogramming through the ectopic expression of defined transcription factors8 — in this Review referred to simply as 'reprogramming' — represents a simple way to obtain pluripotent stem-cell lines from almost any somatic tissue and mammalian species. The use of such cells also circumvents the ethical issues associated with human ESCs.

Reprogramming entails the in trans expression in a somatic cell of a set of core pluripotency-related transcription factors (in most cases OCT4 (also known as POU5F1), SOX2, KLF4 and MYC (also known as c-MYC)(OSKM)). When successful, tightly compacted colonies appear on the culture dish; these colonies resemble ESCs morphologically, molecularly and phenotypically9,10,11,12. These induced pluripotent stem cells (iPSCs) are relevant to a range of applications, including: autologous cell therapy; the modelling of monogenic and multigenic diseases; the study of complex genetic traits and allelic variation; and as substrates for drug, toxicity, differentiation and therapeutic screens. To serve these various purposes, a multitude of protocols for iPSC generation have been developed in recent years. They use, for example, different mouse13,14,15,16 and human donor populations17,18,19, or vary the number, identity and delivery mode of the reprogramming factors20,21,22.

iPSCs represent a widely available, non-controversial and practically infinite source of pluripotent cells. Unlike human ESCs, their usage is not restricted, so most laboratories can now develop research programmes using human pluripotent stem-cell lines. However, one needs to choose a strategy to obtain iPSCs that is suited to the research aims. The simplest approach is to obtain an existing line from another laboratory, but there are also now many options available for generating them in-house. The scope of this Review is to provide an overview of these methodologies and the way they influence the ease, efficiency or kinetics of reprogramming, as well as their expected effects on the genome, epigenome and transcriptome of the pluripotent lines generated. We will see that, as with choosing the appropriate menu for a specific diet, different reprogramming strategies are appropriate for different studies, with the correct approach depending on the priorities of the specific application for which the cells are to be used (Table 1).

Table 1 Considering reprogramming in the light of downstream applications

Because the iPSC field has been extremely prolific during the past few years, it is beyond the scope of this Review to give an exhaustive list of all existing approaches. Supplementary information S1–S4 (tables) provide more details, and we provide a comprehensive database of reprogramming experiments in human and mouse cells at

On the variability of reprogramming

Although direct reprogramming is conceptually and technically simple, it is an extremely slow and inefficient process influenced by several variables that affect its efficiency, reproducibility and the quality of the resulting iPSCs. Before choosing a reprogramming approach it is therefore important to identify these variables. Depending on the application, the appropriate protocol will not only have to take into account the efficiency but also the reproducibility or the quality of the reprogrammed cells. Although this is conceptually straightforward, there is as yet no clear consensus on how to properly measure reprogramming efficiencies (Box 1), reproducibility or iPSC quality (Box 2), making it difficult to properly evaluate these parameters23,24,25. Despite this caveat, some guidelines can be extracted from the literature, thereby allowing an estimate of the effect of these variables on reprogramming

The donor cell type

Reprogramming requires the delivery of certain factors into a specific cell type and their adequate expression under defined culture conditions for a period of time, which varies depending on the cell type, species and delivery method. Depending on the donor cell type, reprogramming is achieved with different efficiencies and kinetics. For example, 8–12 days are required to reprogramme mouse embryonic fibroblasts (MEFs) using retroviruses, whereas the same process takes 20–25 days for human foreskin fibroblasts (HFFs). So far, fibroblasts remain the most popular donor cell type, and were used in more than 80% of all reprogramming experiments published. As a result, several studies have analysed the reprogramming capacity of alternative cell types that are of particular interest owing to their ease of reprogramming (Fig. 1), availability or therapeutic relevance. Compared with fibroblasts, human primary keratinocytes transduced with OSKM reprogramme 100 times more efficiently and twofold faster. Moreover, these cells can be obtained simply by culturing a plucked hair17. Alternatively, cord blood CD133+ cells require only OCT4 and SOX2 to generate iPSCs. In theory, their availability through cell banks could offer a logistic advantage over the use of other adult somatic cell types for the purpose of creating iPSC banks covering a consistent range of haplotypes18.

Figure 1: The reprogramming menu.
figure 1

Any reprogramming experiment is determined by a number of preliminary choices regarding the donor cell type to reprogramme, the factors to use and the mode of their delivery. These choices depend not only on the availability of the cells but also on the purpose that the reprogrammed cells will serve. This figure shows how cell type, reprogramming factors and delivery method might each be evaluated and chosen. Information on the factors that are not commented on directly in the main text can be found using this online database: CHD1, chromodomain-helicase-DNA-binding protein 1; DNMT1, DNA methyltransferase 1; DPPA4,developmental pluripotency associated 4; E-cadherin, epithelial cadherin; ERK, extracellular signal-regulated kinase; MAPK, mitogen-activated protein kinase; MMLV, Moloney murine leukaemia virus; PRC2, Polycomb repressive complex 2; SV40LT, SV40 large T antigen; TERT, telomerase reverse transcriptase; TGFβ, transforming growth factor-β.

The increase in reprogramming efficiency and/or decrease in factor requirement of specific donor populations are attributed to high endogenous levels of certain reprogramming factors — which obviates their expression in trans — and/or intrinsic epigenetic states that are more amenable to reprogramming. The first hypothesis is supported by the fact that neural progenitor cells, which express SOX2 endogenously, reprogramme in the absence of exogenous SOX2 (Refs 15, 16) or with OCT4 alone26.

The differentiation status of the starting cell type also affects reprogramming efficiency. For example, haematopoietic stem and progenitor cells generate 300 times more iPSC colonies than do terminally differentiated B and T cells27. The differences in reprogramming among cell types are not restricted only to the efficiency, but can also affect the quality of the iPSCs. For instance, iPSCs derived from mouse tail-tip fibroblasts have a higher tendency to form teratomas than do those derived from MEFs or hepatocytes28. The choice of cell type is therefore an important aspect to consider before starting any experiment. It will usually depend on cell availability and will affect the requirement for ectopic factors, the efficiency and kinetics of reprogramming, and the quality of the resulting iPSCs.

The reprogramming cocktail

Pluripotency. After choosing a starting cell type, one needs to select a cocktail of reprogramming factors (Fig. 1) and, if required, facilitating compounds. Many of the factors that induce reprogramming are genes that are normally expressed early during development and are involved in the maintenance of the pluripotent potential of a subset of cells that will constitute the ICM of the pre-implantation embryo and, later, the embryo proper. This is the case for OCT4, SOX2 and NANOG, which are core pluripotency transcription factors. When NANOG is expressed along with OSKM in mouse B cells, the time until colony appearance is reduced by half compared with that taken by OSKM alone29. When UTF1, another pluripotency transcription factor, is expressed with OSKM in human primary fibroblasts, more colonies with high levels of alkaline phosphatase are generated30. Similarly, when compared with OSK alone, the overall number of iPSC colonies is increased tenfold when the transcription factor SALL4 (which has been associated with pluripotency) is co-expressed in human fibroblasts31. The ectopic expression of these factors may allow the establishment of an embryonic-like transcriptional cascade that is sustained and stabilized by the reactivation of the endogenous core pluripotency network.

Cell proliferation. Other factors, such as MYC and KLF4, directly or indirectly affect cell proliferation. Telomerase reverse transcriptase (TERT) and the SV40 large T antigen (SV40LT), two proteins that have positive effects on proliferation, increase the appearance of ESC-like colonies when combined with OSKM32. The influence of cell-cycle regulators on reprogramming has also been highlighted using chemical compounds. Specific inhibition of the mitogen-activated protein kinase kinase (also known as MEK) signalling using a compound (PD0325901) increases the number of fully reprogrammed colonies obtained from neural precursor cells infected with Oct4 and Klf4 (Ref. 33). MicroRNAs (miRNAs) are also known to influence pluripotency and reprogramming34, and some miRNAs from the miR-290 cluster — called the ESC-specific cell-cycle regulating (ESCC) miRNAs — contribute to the unique cell cycle of ESCs35. The introduction of OSK plus miR-291-3p, miR-294 or miR-295 into Oct4–GFP reporter MEFs increases the number of GFP+ colonies compared with OSK alone. miR-294 has the most marked effect, increasing the efficiency of reprogramming tenfold. These ESCC miRNAs are believed to be downstream effectors of MYC and show clear potential to enhance the production of mouse iPSCs36. Finally, some factors inhibit reprogramming barriers, such as senescence and apoptosis, and allow an increase in both the speed and efficiency of reprogramming; for example, in mouse cells, inhibition of P53 or members of its pathway using short hairpin RNAs or knockout alleles has this effect30,37,38,39,40,41.

Epigenetics. Chromatin remodelling is a rate-limiting step during iPSC generation42, and chemical compounds that alter DNA methylation or chromatin modifications improve reprogramming in various cell types. Treatment with the DNA methyltransferase inhibitor 5′-azacytidine or histone deacetylase (HDAC) inhibitors (such as hydroxamic acid (SAHA), trichostatin A (TSA) and valproic acid (VPA)) improves reprogramming in MEFs42. By combining the glycogen synthase kinase 3 inhibitor CHIR99021 with tranylcypromine (Parnate) — an inhibitor of lysine-specific demethylase 1 — human primary keratinocytes reprogramme with only OCT4 and KLF4 (Ref. 43). Moreover, VPA enables the induction of pluripotency in neonatal HFFs and dermal fibroblasts with OCT4 and SOX2 alone44. During embryonic development, the G9a histone methyltransferase mediates the epigenetic repression of Oct4 (Ref. 45), which might explain why an inhibitor of G9a (BIX-01294) allows reprogramming of MEFs with only OCT4 and KLF4 (Ref. 46). Butyrate also affects histone H3 acetylation and promoter DNA demethylation, and alters the expression of endogenous pluripotency-associated genes, including developmental pluripotency associated 2 (DPPA2)47. Vitamin C also significantly improves the reprogramming of MEFs and adult mammary gland fibroblasts, in part by alleviating cell senescence48 and inducing DNA demethylation49.

The culture conditions

After deciding on the combination of factors that are best suited to a specific cell type, one needs to consider the conditions in which the cells will undergo reprogramming. For example, culture conditions, supportive cells and medium composition are all parameters that have been shown to modulate reprogramming efficiencies. Reprogramming under hypoxic conditions of 5% O2 (similar to those found in some stem-cell niches, such as the bone marrow), instead of the atmospheric 21% O2, increases the reprogramming efficiency of mouse and human cells by 40- and fourfold, respectively. When combined with VPA, the efficiency increases to 200-fold in mouse cells50. Supportive feeder cells secrete growth factors that are required for ESC survival and/or proliferation and inhibition of ESC spontaneous differentiation51. Marson et al.52 have shown that adding medium conditioned by cells expressing WNT3a promotes the generation of iPSCs in the absence of MYC. Moreover, by testing different culture conditions, Okada et al.53 found that serum-free medium (KK20) allows iPSCs to be obtained at an earlier time point.

Therefore, the questions that need to be considered before a specific reprogramming method is selected are: which cell type, which factors and which culture conditions should be used? Although multiple combinations are possible, in most cases the reprogramming of MEFs and neonatal human dermal fibroblasts is accomplished using OSKM and ESC culture conditions. Sometimes, MYC is eliminated or substituted by other factors because of the oncogenic risk associated with this gene. Additional factors or compounds that could improve the quality of the reprogrammed cells are sometimes added to the OSKM set, although their use should be properly assessed in each situation. Below, we describe the reprogramming methods currently available. We hope that the overview of considerations and options presented above will help readers to understand why some methods are better suited to some applications and which hurdles the protocol modifications try to overcome (Fig. 1; Table 1). Reprogramming methods can be divided into two classes, those involving the integration of exogenous genetic material and those involving no genetic modification of the donor cells.

Integrative delivery systems

Viral delivery systems. The delivery of the OSKM transcription factors into mouse or human fibroblasts was originally achieved using moloney murine leukaemia virus (MMLV)-derived retroviruses (Fig. 2) such as pMXs54,55,56, pLib12 or pMSCV17,57. These vectors have cloning capacities of around 8 kb, allow delivery of genes into the genome of dividing cells and are usually silenced in immature cells such as ESCs58,59. Silencing is important because only an iPSC that has upregulated the endogenous pluripotency gene network and downregulated the expression of the transgenes can really be considered to be fully reprogrammed60. The vector, in which the reprogramming cDNA is cloned, provides a viral packaging signal, as well as transcription and processing elements. On transfection into a packaging cell line that expresses a specific viral envelope protein (which determines the range of cell types that can be infected), high titre, replication-defective viruses are produced; they can infect donor cells with efficiencies of up to 90%. The efficiency of iPSC generation using MMLV-derived retroviruses expressing each gene in the OSKM set separately is ~;0.1% in mouse embryonic fibroblasts and ~0.01% in human fibroblasts.

Figure 2: Viral delivery methods.
figure 2

A flow diagram summarizing the main viral delivery methods, with their advantages and caveats shown below. For each of the methods, the design of the vector is shown at the top, followed by the status of the cell after initial delivery of the vector. The blue cells show the status of the vector in reprogrammed cells (induced pluripotent stem cells). The bottom level (orange cells) shows what might happen to the transgenes after the pluripotent cell is differentiated. The methods are described in more detail in the text.DOX, doxycyclin; MMLV, Moloney murine leukaemia virus.

Lentiviral delivery vectors (Fig. 2) have also been successfully used to express different sets of reprogramming factors in somatic cells22,61. They are generally derived from HIV, exhibit slightly higher cloning capacities (8–10 kb) and usually have higher infection efficiency than MMLV-based retroviruses. Moreover, they allow infection of both dividing and non-dividing cells. The efficiency of reprogramming using lentiviral vectors is comparable to that with MMLV-derived retroviruses. Compared with MMLV-derived vectors, lentiviruses are less effectively repressed in pluripotent stem cells62; this can complicate the identification of bona fide iPSC clones60. Although this issue could not be addressed using constitutive lentiviral vectors22,61, tet-inducible reprogramming lentiviruses allow expression of the reprogramming factors in a controllable manner. Although their preparation is slightly more complicated and time-consuming than that for MMLV-derived retroviruses, the main advantage they present is their availability as inducible systems.

Although they are efficient and reproducible, reprogramming using viruses entails the production of potentially harmful viral particles that express potent oncogenes such as MYC. iPSC lines generated using these vectors carry randomly distributed viral transgene insertions63, which could disrupt the expression of tumour suppressor genes if there are insertions in the open reading frames or alter the expression of oncogenes if inserted nearby. Moreover, they unavoidably generate heterogeneous iPSC lines, which could complicate comparative analysis. Even if properly silenced, viral transgenes can eventually be reactivated during differentiation or during the life of iPSC-derived or transplanted animals, leading to tumours11. These tumours result either from basal expression levels of the MYC transgenes or other oncogene-related factors (if present in the reprogramming set) or tissue-specific reactivation of these transgenes owing to promoter- or enhancer-trapping events. The use of Cre-deletable64 or inducible lentiviruses has solved some of these problems65, but viral systems still lack the safety required for therapeutic applications.

Transfection of linear DNA. If aiming to avoid the use of viral vectors, standard DNA transfection using liposomes or electroporation is a good alternative (Fig. 3). Compared with viruses, however, transduction efficiency is much lower; substantially fewer donor cells receive the full set of reprogramming factors. A crucial improvement has been the design of polycistronic vectors that allow the expression of several cDNAs from the same promoter. These constructs include self-deleting 2A peptide sequences (~20 amino acids long) from the foot-and-mouth disease virus (FMDV) or other picornaviruses66,67. When cloned in between different cDNAs, 2A peptide sequences allow ribosomes to continue translating the downstream cistron after releasing the first protein with its carboxyl terminus fused to 2A. This results in the expression of almost stochiometric amounts of each protein encoded by the polycistron. Such a system has been successfully tested in ESCs68.

Figure 3: Non-viral delivery methods.
figure 3

A flow diagram summarizing the main non-viral delivery methods, with their advantages and caveats shown below. DNA-based delivery methods include those that do or do not involve integration into the genome. For each of the methods, the design of the vector is shown at the top, followed by the status of the cell after initial delivery of the vector. The coloured bars represent the transgenes. The blue cells show the status of the vector in reprogrammed cells (induced pluripotent stem cells). The bottom level (orange cells) shows cells after differentiation — in each case the cells should be transgene-free. The methods are described in more detail in the text. The use of small molecules that accelerate or replace the action of reprogramming factors is not included in this figure; these could be added to any delivery method described in this figure or in Fig. 2. PB, piggyBac; oriP/EBNA1, oriP/Epstein–Barr nuclear antigen-1-based episomal vector.

Using a linearized 2A-peptide-based polycistronic vector flanked by loxP sites, Kaji and colleagues successfully reprogrammed mouse fibroblasts. Approximately 10% of the lines they generated showed single insertions of the construct, indicating that single-copy polycistronic OSKM expression cassettes are sufficient to achieve direct reprogramming. By transiently expressing the Cre recombinase, they then induced recombination between the loxP sites to delete the reprogramming construct69. Such a system is appealing for its simplicity; however, owing to the low percentage of cells transfected with the reprogramming construct and the inherent low efficiency of reprogramming, it requires a large number of donor cells, which may be difficult to obtain for certain cell types. Moreover, obtaining transgene-deleted iPSCs is not necessarily straightforward because many of the colonies with the deletion start to differentiate; this indicates that many clones obtained using this system represent reprogramming intermediates69. The main advantage of this approach is the possibility of deleting the reprogramming cDNAs in iPSCs, which would improve their differentiation potential and, perhaps more importantly, avoid the reactivation or constitutive expression of the reprogramming factors, thus, in theory, reducing their oncogenic potential.

The observation that polycistronic vectors allow reprogramming of somatic cells through a single insertion also encouraged some researchers to include them in integrative MMLV-derived retroviral vectors70 and lentiviral vectors71,72 (Fig. 2), substantially reducing the number of genomic insertions compared with single-factor-expressing viruses. By including loxP sites, such vectors represent an easy way to induce transgene-free iPSCs from various donor sources with higher transduction efficiencies than naked DNA73. These vectors eliminate the oncogenic risk related to transgene reactivation and have a positive effect on the differentiation potential of the resulting iPSCs73.

piggyBac transposon. To enhance the stable integration of non-viral constructs, Kaji and others moved to vectors based on the piggyBac (PB) transposon74 (Fig. 3). The PB transposase is active in mouse75 and human ESCs76, and mediates a higher genome integration efficiency than random integration of linearized plasmids. The reprogramming system includes the PB transposase that mediates gene transfer and a transposon containing the sequence of interest flanked by the 5′ and 3′ terminal repeats required for transposition74,77,78. The PB system is usually composed of a donor plasmid containing the transposon, co-transfected with a helper plasmid expressing the transposase74,75,76,79. Cre-excisable linear transgenes leave a genomic scar, including the loxP site, after Cre deletion, whereas PBs are, in theory, precisely deleted without modifying the sequence of the integration site upon remobilization by the transposase. Using PB-based reprogramming vectors, a number of groups have induced mouse and human pluripotent stem cells from fibroblasts and subsequently deleted the transgenes80,81, thus leading to theoretically genetically unmodified iPSCs. Among integrative methodologies, this approach is the only one that guarantees a precise deletion of the transgenes, although alterations are sometimes observed in the insertion sites, which therefore need to be sequence-verified32.

In Cre or FLP recombination, recombination in cis between the target sites is highly favoured compared with recombination in trans, which leads to a unidirectional reaction and loss by dilution of the circular deleted fragment; by contrast, the PB transposase promotes deletion and integration at similar efficiencies, allowing the transposon to 'jump' from site to site until the transposase is expressed. As a result, the expression window of the PB transposase needs to be tightly controlled because long exposure times lead to several rounds of excision-integration, increasing the risk of non-conservative deletion. To reduce such risk and facilitate the isolation of cells without the transposon, it is highly recommended that negative selection genes, such as thymidine kinase (Tk), are included in the transposon.

Integrative delivery systems can enable efficient generation of iPSCs with single transgene insertions, which can be deleted after reprogramming. Deletion lowers the risks of insertional mutagenesis or oncogenesis (by precluding MYC reactivation) and improving the differentiation capacity of iPSCs (preventing basal expression of the core pluripotency reprogramming transcription factors OCT4, SOX2 or NANOG). Although these are major improvements in terms of the safety and quality of iPSCs, their possible effects during the reprogramming process in terms of genomic stability or possible aberrant epigenetic remodelling still need to be evaluated.

Non-integrative approaches

Non-integrative approaches address a major limitation of iPSCs: the permanent genetic modification resulting from the integration of classic retroviral or lentiviral vectors, or the genomic scars left behind by deletion of Cre-deletable viral vectors, naked DNA transgenes or non-conservative transposon remobilization. The different approaches that are currently available can be subdivided into four main categories: integration-defective viral delivery (Fig. 2), episomal delivery, RNA delivery and protein delivery (Fig. 3). Although the kinetics of reprogramming vary between different starting cell types and species, the generation of stable iPSCs usually requires several weeks to complete. Depending on the starting population, some of the non-integrative approaches are difficult to apply owing to poor infection or transfection efficiencies, poor cell survival, long reprogramming kinetics or other limitations. These considerations underline one of the major drawbacks of these methodologies: they are usually inefficient and poorly reproducible, which is the principal reason why no consensus has yet been reached in the community regarding a method of choice.

Integration-defective viral delivery. One of the first attempts to generate integration-free iPSCs was reported by Stadtfeld et al.82, who used a replication-defective adenoviral vector, pHIHG-Ad2. The authors cloned the OSKM set as single factors into pHIHG-Ad2 and were able to generate transgene-free iPSCs after infection of mouse hepatocytes with adenoviral particles. However, for mouse fetal liver and adult fibroblasts, the authors were able to obtain transgene-free iPSCs only when the vectors carrying Sox2, Klf4 and Myc were complemented in trans by a stably integrated inducible Oct4 transgene82, owing to low infection efficiencies or the transcriptional status of these donor cells. The authors also identified several tetraploid iPSC clones derived from mouse fetal liver cells and mouse adult hepatocytes, probably reflecting the level of endogenous polyploidy of the liver83,84. Using similar vectors, Zhou et al. generated diploid transgene-free iPSC lines from human fetal fibroblasts85. The efficiency of iPSC generation using this system in the mouse ranges between 0.0001% and 0.0018%, which is approximately three orders of magnitude lower than that for retroviruses.

Human fibroblasts and terminally differentiated circulating T cells have also been successfully reprogrammed using F-deficient sendai viral vectors86,87. These vectors efficiently transfer foreign genes into a wide range of host cells88 and replicate in the form of negative-sense ssRNA in the cytoplasm of infected cells89,90. Using Sendai viral vectors expressing OSKM, Fusaki et al.86 generated iPSC lines and were able to isolate a few clones that showed no presence of viral RNA. Although an appealing method, the viral RNA replicase of these vectors is extremely sensitive to transgene sequence content. Furthermore, because these viral vectors replicate constitutively, they are difficult to eliminate from host cells, making it challenging to properly isolate transgene-free clones, even at high passage numbers86. High passage numbers also increase the probability of generating aneuploid iPSC lines owing to longer exposures to Myc.

Transient episomal delivery. As an alternative to integration-defective viruses, some authors have developed reprogramming strategies based on direct delivery of non-replicating91,92,93 or replicating episomal vectors94. These methods are appealing because they are relatively simple to implement with a standard laboratory set-up and molecular biology experience, avoiding the time-consuming and labour-intensive production of viral particles.

By serial transfection of two plasmids expressing, respectively, OSK and Myc91,95, or a single plasmid expressing the full OSKM set as a polycistron92, iPSCs that showed no sign of plasmid integration were obtained from MEFs. Using such methods, only a low percentage (for example, 33%91 and 8%92) of the iPSC lines generated are free of plasmid integration (Fig. 3). There are several possible reasons for this, including: low transfection efficiencies for large plasmids (5–10 kb) that result in few cells receiving the appropriate dose of plasmid over the full reprogramming period; premature dilution of the vectors in actively proliferating cells; or the active silencing of prokaryotic sequences contained in the backbone of these vectors in mammalian cells, leading to downregulation of the reprogramming factors96. These reasons probably explain this method's failure to produce iPSCs from HFFs or keratinocytes, because these cell types require sustained expression of OSKM for a longer duration than do MEFs to reach pluripotency (J.C.I.B. and F.G., unpublished data).

To circumvent the need for serial transfection and to solve the problem of episome dilution through cell division, Yu and colleagues used oriP/Epstein–Barr nuclear antigen-1-based episomal vectors (oriP/EBNA1)94 (Fig. 3). oriP/EBNA1 vectors are maintained through cell division and under selection conditions as stable extra-chromosomal replicons that require only a cis-acting oriP element97, a trans-acting EBNA1 gene and a positive selection gene98. These vectors can be transduced into donor cells using standard transfection procedures and can be removed by culturing the cells in the absence of drug selection. By co-transfecting three oriP/EBNA1 vectors expressing respectively, OCT4SOX2NANOGKLF4, OCT4SOX2SV40LTKLF4 and MYC LIN28 , and in the absence of any drug selection, Yu and colleagues successfully generated iPSC colonies from HFFs. Analysis of the derived subclones revealed that one-third of them were devoid of plasmid DNA. The reprogramming efficiency of human fibroblasts using oriP/EBNA1 vectors is, however, extremely low (3 to 6 colonies per million cells nucleofected), which may reflect the low transfection efficiency of such large plasmids (more than 12 kb), their gradual loss through cell division in the absence of drug selection or active silencing through DNA methylation, resulting in low levels of expression of the reprogramming factors. A major concern about this system is the use of the SV40LT antigen as one of the reprogramming factors. Because this potent viral oncoprotein is able to inactivate both the p53 and the retinoblastoma pathways, the result could be the generation of iPSC lines with higher tumorigenic potential. This aspect still needs to be properly addressed.

In order to decrease the size of the reprogramming episomes and delete potentially methylatable prokaryotic backbone sequences, minicircle vectors represent an interesting solution that allows the expression of the reprogramming factors as non-integrating, non-replicating episomes93. These vectors are supercoiled DNA molecules that lack a bacterial origin of replication and antibiotic resistance gene because their backbone is removed by PhiC31-mediated intramolecular recombination before purification99,100. Compared with plasmids, minicircle vectors show higher transfection efficiencies (their size is usually reduced by at least by 3 kb, the average size of the backbones usually found in episomal vectors) and longer ectopic expression of the transgenes due to lower activation of exogenous DNA-silencing mechanisms99,100. By cloning a 2A-peptide-based polycistronic cassette including OCT4, SOX2, LIN28 and NANOG (OSLN), plus a GFP reporter gene in a single minicircle vector, Jia and colleagues93 reprogrammed human adipose stem cells in 14–16 days with an average efficiency of ~0.005%. Subsequent Southern blot analysis suggested that none of these iPSC lines carried integration of the minicircle vectors, although more sensitive assays for integration, such as PCR, were not performed93.

RNA delivery. In order to completely eliminate plasmid or viral vectors, Warren et al.101 developed a system that achieves the efficient conversion of different human somatic donor cells into iPSCs using direct delivery of synthetic mRNAs (Fig. 3). The efficiency reached with this approach is much higher than that achieved with other non-integrative systems, with 2% of neonatal fibroblasts being converted into iPSCs in just 17 days. This system requires modification of in vitro transcribed RNAs in order for them to escape the endogenous antiviral cell defence response to ssRNA. Phosphatase treatment, incorporation of modified ribonucleoside bases substituting 5-methylcytidine for cytidine and pseudouridine for uridine, combined with the addition of a recombinant version of B18R protein in the medium, allowed for high, dose-dependent levels of protein expression with high cell viability. By delivering synthetic RNAs encoding OSKM and Lin28, reprogramming was achieved by serial transfection of different donor populations using a cationic vehicle101. Although this system is extremely appealing for its simplicity and efficiency, the high gene dosages of the reprogramming factors resulting from direct mRNA delivery may represent an oncogeneic risk owing to higher expression levels of Myc affecting genomic stability. A similar delivery method that avoids this potent oncogene would represent an improvement, although a direct assessment of the mutation load of iPSCs generated using this approach will be required, as for any of the approaches described here.

Protein delivery. Another way to avoid the introduction of exogenous genetic material into donor cells is to deliver the reprogramming factors as proteins (Fig. 3). Several studies have demonstrated that proteins can be delivered directly into cells in vitro and in vivo when fused with peptides mediating their transduction, such as HIV transactivator of transcription (Tat) and poly-arginine102,103,104. Using this approach, Zhou et al. generated recombinant OSKM proteins fused with a poly-arginine transduction domain. They expressed these engineered proteins in Escherichia coli in inclusion bodies; the proteins were then solubilized, refolded and further purified. By serial transduction of Oct4GFP reporter MEFs with OSKM or OSK recombinant proteins, the authors obtained GFP+ colonies if the HDAC inhibitor VPA was also added to the media105. Similarly, Kim and co-workers fused each of the OSKM factors with a myc tag and a tract composed of nine arginines. They generated four stable HEK293 cell lines expressing each of the four human recombinant reprogramming factors and applied the extracts of these cells to human neonatal fibroblasts for 8 hours per week, for a total of 6 weeks. They were able to obtain iPSC colonies after dissociation and replating on MEF feeder cells106. Although promising, protein-based strategies show extremely slow kinetics and poor efficiencies. Moreover, the recombinant proteins used in these approaches are usually difficult to reproducibly purify in the required amounts, making them difficult to use routinely in the laboratory.

Strategic choices and future directions

At present in the iPSC field, it is still difficult to choose a reprogramming strategy that is fitting for all purposes. This can be illustrated by considering two different research programmes: the first focused on deciphering the mechanisms of reprogramming, the second intended to generate clinically relevant iPSCs. In the first situation, the reprogramming approach needs to be robust and efficient; a delivery method and a combination of factors that can achieve this is required, regardless of the presence of genomic modifications. For example, using integrative inducible lentiviruses and subsequently deriving secondary iPSCs will meet these requirements. By contrast, the second case requires a non-integrative or semi-integrative approach in order to avoid or control genomic modifications. Multiple methods have been proposed, but all are inefficient. An RNA-based approach published recently101 seems promising owing to the high efficiency achieved. In any case, the choice of starting-cell population will depend on the availability of the cell type, the ease with which it is reprogrammed and the efficiency it yields. Importantly, the use of 'safe' approaches does not necessarily prevent variability in the expression levels of lineage-specification genes or the occurrence of aberrant epigenetic remodelling, which may limit downstream applications of iPSC technology.

The improvement in reprogramming efficiency and/or kinetics that can be achieved with small molecules makes them an attractive avenue for further research, although they must be treated with caution because some can be tumorigenic (for recent reviews, see Refs 107, 108). Small molecules represent a powerful alternative or support for reprogramming because they can target different cellular pathways that control cell fate, state and function, but their specificity is sometimes difficult to assess. Their progressive introduction in reprogramming protocols and/or implementation of large-scale screens will probably be key to identifying new pathways that might allow the replacement of current reprogramming factors or that might have a positive effect on iPSC generation.

Since the first published demonstration that fibroblasts can be reprogrammed by retroviral delivery of just four factors (OSKM), a substantial number of alternative approaches have been developed to induce pluripotency in somatic cells. To properly assess the improvement that each of the methods provides and to give a more precise idea of their real contribution to reprogramming, it will be crucial to test them using commonly accepted standards. In addition to the use of oncogenes in reprogramming cocktails and the issue of viral integration, reprogramming itself may have an effect on a cell's genome, especially given that the process takes many weeks and is rather inefficient. The low efficiency and slow kinetics might subject cells to detrimental alterations, such as the accumulation and/or selection of subtle genetic and epigenetic abnormalities before or during reprogramming, which could favour the activation of growth pathways and the inhibition of tumour suppressor pathways (F.G. and J.C.I.B., unpublished data). A crucial challenge in the iPSC field will be to properly determine how these various methodologies affect the quality of iPSCs in terms of transcriptional signatures, epigenetic status, genomic integrity, stability, differentiation and tumour potential. Whole-genome sequencing platforms will probably play an important part in the future in assessing the integrity of the genome of iPSCs and will certainly improve our understanding of the mechanism by which reprogramming occurs in a specific cell type.