Disease modelling

Following the first human fibroblast derivation of iPSCs, investigators went on to show that the reprogramming protocol could be used in cells taken from a wide range of mendelian and complex genetic disease backgrounds (Ref. Reference Park22). This observation was then followed by studies proving that human iPSCs could be differentiated into cells similar to adult somatic cells. A range of different human cells with varying relevant functional properties have thus far been generated, including motor and dopaminergic neurons (Refs Reference Dimos23, Reference Soldner24, Reference Karumbayaram25), cardiac cells (Ref. Reference Zhang26), pancreatic cells (Ref. Reference Zhang27), adipocytes (Ref. Reference Taura28), blood cells (Ref. Reference Choi29), vascular cells (Ref. Reference Taura30), retinal progenitors (Ref. Reference Lamba31) and liver cells (Refs Reference Touboul32, Reference Si-Tayeb33, Reference Sullivan34, Reference Gallicano and Mishra35).

However, it was not clear whether the derived cells would be of any ‘clinical’ relevance until publication of two recent landmark studies. In the first study (Ref. Reference Ebert36), investigators described disease modelling of a patient suffering from spinal muscular atrophy (SMA). This study showed that the iPSC-derived neural lineage undergoes selective neuronal death, which could be reversed upon addition of compounds known to increase production of survival motor neuron protein (SMN). However, the authors were not able to produce fully functional neurons displaying the disease phenotype. As a result, they were unable to show that SMN protein was itself reduced in the adult cell type (although a decreased amount of mRNA transcripts was shown), thus making the model an excellent proof-of-concept tool for the production of an in vitro iPSC-based SMA model, but an incomplete disease model in itself. In the second study (Ref. Reference Lee37), the authors describe a patient-specific iPSC-based model of familial dysautonomia. The model impressively showed faithful replication of three key pathogenic features of the disease: (1) defects in gene (IKBKAP) splicing, (2) decreased neurogenesis and (3) defective migration of iPSC-derived neural precursors. However, as with the first study, no evidence was provided to show that decreased protein production was replicated in the fully mature adult cell, nor that lack of protein resulted in adult cell degeneration or paucity of production. Both studies were restricted to the neural lineage, dealt with very small numbers of patients and concerned very rare diseases of developmental abnormality. Furthermore, neither of these models proved that conservation of disease pathology occurring at a protein structural level was possible using human iPSC technology. It is therefore not yet clear whether the myriad of subtle interactions responsible for the conservation of protein translation and trafficking will be preserved following the stressful conditions exerted by ex vivo reprogramming and cellular differentiation upon human iPSCs.

It is significant to recollect at this point that although human ES cells have been studied for over 10 years, we are still unable to reliably achieve fully mature, completely functional cells of interest in vitro. Furthermore, to assume that generation of just one particular cell type will faithfully recapitulate a disease process in vitro is too simplistic. Investigators will need to invest time and resources to establish which cell types are fundamental to the disease of interest and then study how best to aggregate these chosen cells into a novel in vitro platform if new insights into disease processes are to be unearthed. Such efforts will surely be worthwhile because human-iPSC-based platforms offer unique opportunities for patient-specific research that are unachievable with currently available cell lines.

Regenerative medicine

No human trials have thus far been conducted using iPSCs. Yet this is not altogether surprising when we consider that in spite of their much longer history, human ESCs have not yet been approved by the United States Food and Drug Administration (FDA) for use in human clinical trials. However, a Phase I clinical trial based on the use of human-ESC-derived oligodendroglial progenitor cells (GRNOPC1) in subacute thoracic spinal-cord injuries (Ref. Reference Zhang, Denham and Thies38) is being planned and should start shortly. Such advancements therefore suggest it will only be a matter of time before iPSC-based trials are also conducted in regenerative medicine. Indeed, several key studies in the mouse have already established the proof of principal that human iPSCs could be used advantageously in cell-based therapy for degenerative disorders affecting a range of organs, including the eye, the heart and the brain. For example, it has been shown that isolation of a purified human-iPSC-derived photoreceptor population followed by transplantation into a normal mouse retina was capable of anatomical integration and expression of photoreceptor markers (Ref. Reference Lamba31). Whereas in the cardiac field, intramyocardial delivery of human iPSCs within infarcted adult mouse hearts yielded progeny that properly engrafted without disrupting the cytoarchitecture. Moreover, the iPSC treatment restored postischaemic contractile performance, ventricular wall thickness and electric stability, while achieving in situ regeneration of cardiac, smooth muscle and endothelial tissue (Ref. Reference Nelson39). Finally, in the brain, it has recently been shown that neurons derived from reprogrammed mouse fibroblasts could functionally integrate into the foetal brain of rat models of Parkinson disease, which went onto to display improved symptoms 1 month after treatment (Ref. Reference Wernig40).

However, in our opinion, the complex biological interactions that might account for such positive results suggests that this type of study may not be so easily translatable into a human context in the near future. Instead, we believe that cell-based therapies for well-characterised monogenetic disorders that affect a single adult cell type will prove easier, more realisable targets. Such diseases will require expert collaboration between two previously distinct fields of research – stem cells and gene therapy. Encouragingly, some very impressive mouse studies have already set a precedent in this area. The first of these studies used iPSCs derived from a humanised mouse model of sickle-cell anaemia. The sickle haemoglobin allele was corrected by introducing the intact human wild-type β-globin gene (HBB) via homologous recombination. These gene-corrected iPSCs were subsequently used to derive hematopoietic stem and progenitor cells. Transplantation of gene-corrected hematopoietic stem and progenitor cells into three sublethally irradiated autologous recipient mice resulted in robust engraftment and a significant reduction of the sickle-cell haemoglobin as measured by haemoglobin electrophoresis at 4 and 8 weeks after transplantation. Analysis of the peripheral blood of the three transplanted mice revealed increased haemoglobin levels and improvements of red-blood-cell morphology on peripheral blood smears. In addition, an overall improvement of the clinical status of the animals was noted following corrective bone-marrow transplantation (Ref. Reference Hanna41). A similar subsequent report from a different laboratory expanded on this first study by describing how iPSCs derived from mouse-tail-tip fibroblasts could be differentiated into endothelial progenitor cells that express clotting factor VIII, both in vitro and in vivo. The iPSC-derived endothelial progenitor cells were then engrafted into the liver of a haemophilia mouse model and provided phenotypic correction of this bleeding disorder for as long as 3 months (Ref. Reference Xu42).

The most recent work in this field, led by Carlos Belmonte, has however given us the clearest indication that iPSC-based technologies could soon be used within a clinically relevant human context. They showed that following correction of the genetic defect, somatic cells from Fanconi anaemia patients could be reprogrammed to pluripotency to generate patient-specific iPSCs. The corrected Fanconi-anaemia-specific iPSCs could be differentiated into haematopoietic progenitors of the myeloid and erythroid lineages, which were phenotypically normal, and therefore disease free (Ref. Reference Raya43). Unfortunately, the resulting progenitor cells could not engraft into an animal model, which clearly highlights one of the main stumbling blocks of this and the stem-cell field in general: the need to optimise protocols of differentiation to generate homogenous adult-like cells capable of in vivo functionality. Similarly to the ESC field, these early preclinical models will therefore require a great deal of supplementary refinement using a variety of relevant animal models before clinical trials will become feasible. But given that many of these platforms have been set up as a result of the previous work in the ESC field, we could assume that these stumbling blocks will soon be overcome. If they are, then what would be the remaining iPSC-specific barriers to clinical application?

What starting material should be used for iPSC generation?

Fibroblasts were used as the starting material for reprogramming in the first cohort of iPSC papers (Refs Reference Takahashi and Yamanaka5, Reference Okita, Ichisaka and Yamanaka6, Reference Maherali44). Since then, a wide range of cell types have been found to be capable of generating iPSCs. In the mouse, iPSCs have been generated from bone marrow cells (Ref. Reference Takahashi and Yamanaka5), hepatocytes and gastric epithelial cells (Ref. Reference Aoi45), pancreatic cells (Ref. Reference Stadtfeld46), neural stem cells (Ref. Reference Kim47) and B cells (Ref. Reference Hanna48). Whilst in the human, iPSCs have been generated from skin fibroblasts, keratinocytes (Ref. Reference Aasen49), foetal neural stem cells (Ref. Reference Kim50), blood progenitor cells (Ref. Reference Loh51) and just recently CD133⁺ (Ref. Reference Giorgetti52) and CD34⁺ cord blood (Ref. Reference Haase53). Choosing the right starting cell is important because safety concerns have been raised by a report from Yamanaka and co-workers who showed that secondary neurospheres (SNSs) from iPSCs derived from different mouse adult tissues varied substantially in their teratoma-forming propensity (Ref. Reference Miura54). This report might turn out to be functionally related to another recent study, which suggested that different target cells might have intrinsic varied capacities to fully reprogramme to the pluripotent state, although this link has not yet been proved (Ref. Reference Ghosh21).

The cell (tissue) type used as a starting population also appears to have a major impact on the efficiency, as well as safety, of reprogramming. In an elegant set of experiments, Jaenisch and colleagues illustrated the impact of the differentiation state of the cell of origin in the mouse by taking advantage of the highly ordered developmental pathway of sequential DNA rearrangements in heavy- and light-chain immunoglobulin loci during consecutive stages of B-cell maturation (Ref. Reference Hanna48). Although nonterminally differentiated pro- and pre-B cells were readily reprogrammed by stably integrated doxycycline-inducible reprogramming factors under conditions adopted to sustain growth of lymphocytes, terminally differentiated IgM⁺ IgG⁺ spleen-derived B cells failed to reprogramme under identical conditions. However, when these differentiated B cells were ‘sensitised’ by altering their mature B-cell identity, reprogramming did occur. Sensitisation was achieved either by introducing the myeloid transcription factor CCAAT/enhancer-binding protein-α or by specific knockdown of the master regulator of mature B-cell development and function, Pax5. These results indicate that nonterminally and terminally differentiated B cells respond differently to the expression of Oct3/4, Sox2, Klf4 and Myc, but provide definite proof for the direct nuclear reprogramming of terminally differentiated adult cells to pluripotency. However, the same team have gone on to show that that all cells (or at least the immune cells used in this elegant mouse model) might be capable of being reprogrammed and that there are two ways for the cells to reprogramme — one that is dependent on cell division, the other that is not (Ref. Reference Hanna55).

What tools should be used for reprogramming?

After identifying the quartet of transcription factors that could induce pluripotency, Yamanaka had to devise a way of getting the factors into the nucleus of the cell of choice and inducing their expression. He chose a retrovirus system that allowed the infection of a high fraction of cells (nearly 80%) and also facilitated a downregulation of the inserted genes after reprogramming because of the inherent ability of ES cells to silence viral promoters. In addition, the use of retroviruses enabled the exogenous factors to be expressed at high levels for long enough to induce the reprogramming transcriptional network. The genetic transduction with exogenous genes, particularly oncogenes such as MYC and KLF4, and the use of integrating retroviral-delivery systems were identified as clear handicaps of iPSC technology with regard to potential future clinical applications. Previous evidence has highlighted the dangers of manipulating the genome (Ref. Reference Hacein-Bey-Abina56): integrated viral vectors have been implicated in gene-expression abnormalities in human iPSCs (Ref. Reference Soldner24) and tumour formation in iPSC-derived mice (Ref. Reference Maherali57). Therefore, investigators began designing novel ways of delivering the factors, such as nonintegrating viruses, piggyBac transposon systems and most recently small molecules and recombinant proteins.

Transgene-free iPSCs can be generated by two basic approaches. The first strategy uses transient factor delivery with vectors that do not integrate into the genome. These experiments stemmed from the question of whether insertional mutagenesis is itself required for iPSC formation and were carried out with mouse cells by two independent groups using either adenoviral vectors (Refs. Reference Stadtfeld46, Reference Zhou and Freed58) or transfected plasmids (Ref. Reference Okita59). In the case of plasmid vectors, repeated rounds of transfection were required to convert foetal fibroblasts into iPSCs, which often contained permanently integrated transgenic DNA. More recently, human fibroblasts have been reprogrammed using self-replicating and selectable episomes, belonging to the Epstein–Barr-virus-derived family of oriP/EBNA1 vectors. Notably, successful iPSC derivation in this system required the addition of three additional factors (Myc, Klf4 and SV40LT) in combination with an alternative reprogramming cocktail comprising Oct4, Sox2, NANOG and Lin-28 (Ref. Reference Yu60). In general, reprogramming with adenoviral or episomal vectors occurs at frequencies 100-times lower than with integrating viral vectors. This is probably due to the dilution of reprogramming factors with cell division, resulting in suboptimal expression of these factors and thus the failure to reprogramme most cells.

The second strategy to produce transgene-free iPSCs entails genomic integration of reprogramming vectors, followed by their excision with highly specific enzymes upon completion of reprogramming. A classical expression–excision system uses vectors flanked by recognition sites for Cre recombinase (34-nucleotide loxP sites). Cre-loxP-based approaches have been used to reprogramme mouse fibroblasts with episomal vectors (Ref. Reference Kaji61) as well as human somatic cells from individuals with Parkinson disease using lentiviral vectors (Ref. Reference Soldner24). Although the efficiencies of iPSC recovery were higher compared with transient delivery approaches, a potential limitation of this technique is that Cre-mediated excision leaves loxP sites at random locations in the genome, which might disrupt the function and expression of endogenous genes (Ref. Reference Vallier, Alexander and Pedersen15). In addition, Cre recombinase might be able to recognise cryptic loxP sites naturally present in the human genome and thus induce genetic abnormalities (Ref. Reference Vallier, Alexander and Pedersen15). In two recent reports (Refs. Reference Woltjen62, Reference Yusa63), an alternative strategy to generate iPSCs at reasonable efficiency and apparently free of genomic alterations was presented. The novel approach is based on the piggyBac transposon, a mobile genetic element from insects that efficiently integrates into the genome preferentially in transcribed regions of mammalian cells. Notably, piggyBac transposons can be entirely removed when re-exposed to transposase. Both integration and excision are catalysed by the piggyBac transposase, which restores the correct genomic sequence in >90% of excision events (Ref. Reference Wang64). However, this approach relies on transfection methods such as nucleofection, whose efficiency is highly variable between cell lines and thus substantial effort is required for optimisation. Consequently, the piggyBac system might be difficult to apply to a broad number of fibroblast lines, especially in a clinical context. A possible way to overcome this might be to use an RNA virus that does not integrate in the host genome. A recent report has shown that the Sendai virus (SeV), which is thought to carry no risk of altering the host genome, is an efficient alternative solution for safely generating iPSCs (Ref. Reference Fusaki65).

Another way to avoid host DNA integration is to completely avoid the use of viruses altogether and generate iPSCs using chemicals and or small molecules. Several groups have already identified chemicals that can improve efficiency or even replace one or two reprogramming factors during iPSC generation (Refs. Reference Huangfu66, Reference Shi67, Reference Shi68, Reference Ichida69, Reference Mali70). Moreover Ding's group have shown that a chemical approach can dramatically improve (200-fold) the efficiency of viral iPSC generation from human fibroblasts, within 7 days of treatment, thus providing an ideal basis for developing safer, more efficient, non-viral methods for reprogramming human somatic cells (Ref. Reference Lin71).

More excitingly, however, two reports have shown generation of protein-induced pluripotent stem cells (piPSCs) from murine (Ref. Reference Zhou72) and human (Ref. Reference Kim73) fibroblasts using recombinant cell-penetrating reprogramming proteins without the requirement of any genetic manipulation. To generate recombinant proteins that can penetrate the plasma membrane of somatic cells, Ding and co-workers designed and fused a poly-arginine (i.e. 11R) protein transduction domain to the C-terminus of four reprogramming factors: Oct4, Sox2, Klf4 and Myc. The proteins were expressed in E. coli in inclusion bodies, which were then solubilised, refolded and further purified before using a calculated repeated protein transduction protocol to reprogramme murine fibroblasts with the aid of valproic acid. The subsequent study by the Kim group was able to generate human iPSCS that did not require addition of further small molecules. They first generated stable mammalian cell lines (HEK293) that could express each of the four human reprogramming factors (Oct4, Sox2, Klf4 and Myc) and fused them with 9R and the Myc tag, then repeatedly applied these recombinant proteins to the human fibroblast culture. Although this protein-induced reprogramming has huge benefits, it is still severely limited by the inefficiency of the process in comparison with viral programming – taking about twice as long (8 weeks) and generating one-tenth of the number of lines (an efficiency of 0.001% versus 0.01%).