Pluripotent Stem Cells in Regenerative Medicine: Challenges and Recent Progress

Home , Directed differentiation, Regenerative Medicine

REVIEWS

Pluripotent stem cells in regenerative medicine: challenges and recent progress

Viviane Tabar1 and Lorenz Studer1,2 Abstract | After years of incremental progress, several recent studies have succeeded in deriving disease-relevant cell types from human pluripotent stem cell (hPSC) sources. The prospect of an unlimited cell source, combined with promising preclinical data, indicates that hPSC technology may be on the verge of clinical translation. In this Review, we discuss recent progress in directed differentiation, some of the new technologies that have facilitated the success of hPSC therapies and the remaining hurdles on the road towards developing hPSC-based cell therapies.

Directed differentiation Access to unlimited numbers of specific cell types on progress has been made in the area of neuronal lineage A method to control the demand has been a long-standing goal in regenerative specification, which is highly dependent on mimicking differentiation of pluripotent medicine. With the availability of human pluripotent in vitro the early patterning signals that impart axial stem cells into specific cell stem cells (hPSCs) and greatly improved protocols for coordinates during neural development. Both small- types, which is typically their directed differentiation, this prospect could become molecule-based and morphogen-based approaches have achieved by providing cells with extrinsic signals in a a reality for several disease-relevant cell types. Recent been developed to derive specific neuronal subtypes precise temporal sequence advances in the stem cell field indicate that the ‘holy from pluripotent stem cells. However, the replacement that mimicks development. grail’ of directed differentiation (that is, the generation of of nerve cells in traumatic or degenerative disorders of unlimited numbers of authentic and genetically matched the central nervous system (CNS) remains a daunting Morphogen A substance that is active in cell types for cell therapy) could indeed translate into task. Recent strategies for in vivo cell-fate conversion are 1–4 pattern formation and that effective therapies for currently intractable disorders , still at early stages of development but could potentially varies in spatial concentration although new challenges are likely to emerge on the road advance as an alternative approach that bypasses the or activity; cells respond towards such translation in humans. need for cell transplantation (reviewed in REF. 8). differently at different In parallel to the improvement in directed differentia- Over the years, the field of directed differentiation threshold concentrations of morphogens. tion, novel technologies have been developed to assess has used three main strategies to specify neural lineages lineage, fate and function of stem cell-derived cell types from hPSCs. These strategies are embryoid body forma- both in vitro and in vivo. Here, we review some of the tion, co‑culture on neural-inducing feeders and direct recent breakthroughs in directed differentiation and neural induction. Early protocols for embryoid body discuss their implications for cell therapy. The ability formation were based on triggering differentiation of to access patient-specific cells at scale and on demand human embryonic stem cells (hESCs) followed by selec- 6 1Center for Stem Cell Biology are also crucial for human disease modelling and hPSC- tion in serum-free media to enrich for neural lineages . and Department of based drug discovery; these applications are not dis- The development of serum-free embryoid body cultures Neurosurgery, Memorial cussed here but have been the subject of a recent review5. enabled the direct induction of neural lineages from Sloan-Kettering Cancer In this Review, we present recent breakthroughs in hPSCs, and the efficiency of serum-free embryoid body Center, 1275 York Avenue, New York 10065, USA. deriving therapeutically relevant cell types from hPSC formation could be greatly improved in the presence 2Developmental Biology sources, discuss some of the tools that made such of the Rho-associated protein kinase (ROCK) inhibi- Program, Memorial progress possible and highlight crucial next steps and tor compound Y-27632 (REF. 9) that prevents cell death Sloan-Kettering Cancer challenges on the road towards clinical translation. of dissociated hPSCs. Stromal feeder-based cultures Center, 1275 York Avenue, have also been widely used for generating neuroepi- New York 10065, USA. e-mails: [email protected]; Deriving neural cell types thelial cells and specific neural populations, includ- 10 [email protected] Neural cell types were among the first lineages to be ing midbrain dopamine neuron-like cells from hPSCs . doi:10.1038/nrg3563 reliably obtained from hPSC sources6,7. Arguably, most Although the mechanism of neural induction (that is,

82 | FEBRUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

stromal-derived inducing activity) remains unclear and For nearly a decade, studies had shown the derivation the use of feeders would greatly complicate translational of cells with midbrain dopamine neuron-like proper- use, this approach has remained in use because of the ties10,16,17. However, those neurons lacked a subset of fea- robust induction efficiencies and the ability to combine tures, such as expression of the DNA-binding forkhead it with other neural inducing strategies. box protein A2 (FOXA2), and did not engraft efficiently. Direct induction protocols do not require embryoid We recently established a novel protocol for deriving body formation or co‑culture for neural induction. Early neurons from hESCs that is based on a more precise attempts at direct conversion were based on the simple developmental patterning of the cells (FIG. 1a). A key fea- switch of hESC cultures to serum-free culture condi- ture of this protocol is the transition of the cells through tions followed by mechanical isolation of spontaneously a floor plate intermediate stage instead of the neuroepi- appearing neural rosette cultures7. However, the use of thelial intermediate that is used in past attempts. The defined neural inducers, such as inhibitors of trans- floor plate is a transient developmental structure that forming growth factor (TGF) and bone morphogenetic has only been recently implicated in midbrain dopamine Embryoid body dual SMAD inhibition A clump of cells that arises protein (BMP) signalling (that is, neuron development. The resulting floor plate-derived when embryonic cells are (dSMADi)), have greatly enhanced the efficiency and dopamine neurons have genetic, biochemical and phy cultured and differentiated in the speed of neural induction11. A particularly attrac- siological features of authentic midbrain dopamine neu- suspension and that can give tive feature of dSMADi is the synchronized differen- rons, and they have been successfully tested in mouse, rise to cell types from all 1 three germ layers (that is, tiation process that yields a nearly uniform population rat and rhesus monkey models of Parkinson’s disease . A the endoderm, mesoderm of early neural cells within ten days of differentiation. key step was the activation of canonical WNT signalling and ectoderm). The use of precise patterning strategies in combination using a small-molecule inhibitor of glycogen synthase with dSMADi results in protocols for the derivation of kinase (GSK), which is a strategy that has been replicated Stromal many CNS and peripheral nervous system (PNS) line- in several additional studies since then18,19. These data Pertaining to connective tissue that is made up of both cells ages from hPSCs. However, regardless of the specific emphasize the importance of fully defining the identity (such as fibroblasts) and matrix neural induction strategy used, the main challenge over of hESC-derived neurons before implantation. (such as collagen). the past ten years has been to develop protocols that implement in vitro the early patterning events that are Striatal neurons. Huntington’s disease is a currently Dopamine neuron A nerve cell that uses the responsible for creating specific neuronal and glial cell untreatable autosomal dominant neurodegenerative neurotransmitter dopamine; types. Only recently have these strategies been refined disease and is characterized by abnormal movements, those in the midbrain are to a level that is sufficient to contemplate translational cognitive decline and various psychiatric problems. affected in Parkinson’s disease. applications for a subset of neural lineages. Recent pro- The genetic cause of the disease is an expansion of gress for three relevant hPSC-derived neural lineages is CAG repeats within the huntingtin gene (HTT)20. Dual SMAD inhibition (FIG. 1) striatal neurons (dSMADi). The concomitant discussed below . Medium spiny are the most severely inhibition of bone affected cell type in patients with Huntington’s disease. morphogenetic protein and Dopamine neurons. Parkinson’s disease is the second There is a long history of fetal tissue-grafting studies Nodal–activin–transforming most common neurodegenerative disorder and is in Huntington’s disease21 that aimed to replace striatal growth factor-β signalling, which is used to obtain neural characterized by the progressive loss of several neural neurons. Similarly to Parkinson’s disease, decades of cells from human pluripotent cell types in the CNS and PNS. Although the causes of experience from the fetal tissue-grafting studies could stem cell sources. Parkinson’s disease remain unknown to a large extent, pave the way for future hPSC-based strategies in the the specific loss of midbrain dopamine neurons in treatment of Huntington’s disease. A major feature of Floor plate Parkinson’s disease is responsible for most of the motor Huntington’s disease is the need to restore long-distance A transient developmental structure along the midline of symptoms of the disease, and most current drugs for connectivity from the striatum to pallidal targets, 18 the embryo that is important Parkinson’s disease are aimed at restoring dopamine which are located in the globus pallidus. A recent study for brain development. function. Dopamine neuron replacement has there- presents a protocol that yields a large proportion of sur- fore been pursued as a potential therapeutic strategy viving neurons that express DARPP32 (also known as Striatal neurons Neurons that lie in the for many decades. The feasibility of dopamine neuron PPP1R1B, which is a marker for spiny neurons) in vivo striatum, which is an area of replacement has been shown using human fetal tissue in (FIG. 1b). It further provides evidence of efferent con- the brain that is involved in more than 300 patients with Parkinson’s disease world- nectivity and behavioural improvement in a striatal fine movements, emotion wide12. However, despite long-term in vivo dopamine lesion model, which suggests considerable transla- and cognition. neuron survival and dopamine secretion, the clinical tional potential. However, the key factors that led to Globus pallidus benefit of using fetal dopamine neurons was modest, the improved performance remain unclear, as the new A subcortical structure of the and a subset of patients developed troubling side effects protocol did not include a novel strategy but optimized brain that is a major element such as graft-induced dyskinesia13,14. The results of both existing conditions (that is, sonic hedgehog (SHH) acti- of the basal ganglia system. open-label and placebo controlled studies are discussed vation and WNT inhibition)22. Ongoing developmental 15 23 Oligodendrocytes in detail elsewhere . Although the interpretation of studies are geared towards further refining the derivation + One of the three main cell these fetal tissue-grafting studies remains controversial, of DARPP32 striatal neurons. types that make up the brain there is a general agreement that fetal tissue is not a suit- parenchyma, the other two able cell source for developing a clinically competitive Glial cells. The derivation of engraftable glial cells — being neurons and astrocytes. future therapy against available pharmacological and for example, by myelinating oligodendrocytes — pre- They produce myelin, which insulates axons to alter the surgical alternatives, and that stem cell-based therapy sents a different set of challenges, such as protracted conduction properties of must overcome the limitations of fetal tissues to become developmental timing. Although early studies sug- neurons. a meaningful therapeutic option. gested that cells with oligodendroglial properties can

NATURE REVIEWS | GENETICS VOLUME 15 | FEBRUARY 2014 | 83

In vitro In vivo a hESC-derived mDA neurons • 6OHDA lesioned mouse d 0–d 13 d 13–d 25 4–5 months • 6OHDA lesioned rat + Pur (SHH) + PDGF-AA • MPTP lesioned + CHIR (WNT) + BDNF monkey + SB (–TGFβ) + GDNF hESCs or Floor plate Young mDA neurons • mDA neuron survival + LDN (–BMP) + cAMP human iPSCs + TGFβ3 • Behavioral improvement

b hESC-derived striatal neurons

d 0–d 12 d 12–d 26 d 26–d 32 d 32–d 40 4–5 months Ibotenic acid lesioned rat + SHH + VPA + BDNF hESCs Neuro- LGE LGE + GDNF Striatal + IGF epithelium precursors precursors neurons • DARPP32+ neuron survival + cAMP • Behavioral improvement

c Human iPSC-derived glial precursors

d 0–d 19 d 19–d 40 d 40–d 160 9 months Neonatal Shiverer mice + FGF2 + RA + PDGF-AA + Pur (SHH) + IGF • Oligodendrocyte survival Human iPSCs Neuro- + B27 Pre-OPCs + Pur (SHH) OPCs and • Remyelination epithelium + T3 astroglia • Extension of lifespan + NT3

Figure 1 | Generation of therapeutically relevant neural lineages from hPSCs. Schematic diagramsNature Reviews show published | Genetics protocols for the generation of midbrain dopamine (mDA) neurons for the potential treatment of Parkinson’s disease, striatal neurons for the treatment of Huntington’s disease and glial precursors for the treatment of demyelinating disorders. Small molecules and growth factors that are used to direct cell fate are indicated below the arrows; the factors that are induced or inhibited are shown in parentheses. All studies showed robust long-term in vivo survival and functional improvement in at least one relevant animal model of disease. Although protocols for generating mDA neurons are relatively fast and efficient, protocols for generating oligodendrocyte precursor cells (OPCs) can take more than five months of in vitro differentiation. a | To differentiate human pluripotent stem cells (hPSCs) into mDA neurons through a floor plate intermediate1, combined induction of neural floor plate using four small molecules from day 0 to day 13 (d 0–d 13) is followed by neuronal differentiation in the presence of a ‘cocktail’ of growth factors that promote mDA neuron fate. Cells are ready for transplantation into rodent or primate models of Parkinson’s disease by day 25 of differentiation, which corresponds to the stage at which mDA neurons are born but still immature. b | To differentiate human embryonic stem cells (hESCs) into striatal neurons2, neural induction into the neural rosette stage (that is, the neuroepithelium) is followed by exposure to a precise concentration of sonic hedgehog (SHH), which is required to specifically induce striatal precursors that emerge from the lateral ganglionic eminence (LGE) region during development. Floating cultures of LGE precursors are replated and further matured into striatal neurons. By day 40, these neurons are transplanted into a striatal lesion model of Huntington’s disease. DARPP32 is a marker for spiny neurons. c | To differentiate human induced pluripotent stem cells (iPSCs) into glial precursors3, neuroepithelial cells are isolated and treated with factors that promote ventral cell fates, followed by a protracted period of in vitro proliferation and maturation to obtain OPCs that are capable of efficient in vivo myelination. 6OHDA, 6‑hydroxydopamine; BDNF, brain-derived neurotrophic factor; BMP, bone morphogenetic protein; cAMP, cyclic AMP; CHIR, CHIR99021 (a small-molecule inhibitor); FGF2, fibroblast growth factor 2; GDNF, glial cell-derived neurotrophic factor; IGF, insulin- like growth factor; LDN, LDN193189 (a small-molecule inhibitor); MPTP, 1‑methyl‑4‑phenyl‑1,2,3,6‑tetrahydropyridine; NT3, neurotrophin 3; PDGF-AA, platelet-derived growth factor AA; pur, puromycin; RA, retinoic acid; SB, SB431542 (a small-molecule inhibitor); T3, tri-iodothyronine; TGFβ, transforming growth factor-β; VPA, valproic acid.

be rapidly derived from hESCs24, subsequent studies culture upon early exposure to retinoic acid and SHH argued that those early lineage cells may not be truly agonists26. The translational potential of hPSC-derived committed and require months of maturation for oligodendrocyte precursors was best illustrated in a in vivo myelination. Nevertheless, those early glial recent study using glial precursors that were derived precursors have developed into the first hESC-based from a long-term in vitro differentiation of human product to reach clinical trials for treating spinal cord induced pluripotent stem cells (iPSCs), which were injury25. More committed cells from the oligodendro- then grafted into the neonatal CNS of myelin-deficient cyte lineage were obtained following long-term in vitro mice3 (FIG. 1c). Under those conditions the precursors

84 | FEBRUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

In vitro In vivo a hESC-derived cardiomyocytes

d 0–d 4 d 4–d 16 4 weeks Cryo-injury guinea pig heart RPMI + B27 RPMI + Activin A + B27 • Cardiomyocyte survival GCaMP3 + BMP4 Cardiac Immature • Electric coupling hESCs mesoderm cardiomyocytes Percoll • Improved mechanistic function

b hESC-derived pancreatic precursors FACS d 12 CD142 3–12 d 0–d 3 d 3–d 12 weeks Streptozocin- d 12–d 20 treated mice + WNT + RA + Activin A + Cyc hESCs Deﬁnitve Pancreatic In vivo diﬀerentiation to 0.2% FBS + Noggin insulin-positive cells with endoderm (2% FBS) endoderm + B27 in vivo glucose response (1% B27) Endocrine precursors

Figure 2 | Generation of mesodermal and endodermal lineages from hESCs. Schematic diagrams show published protocols for generating cardiomyocytes as an example of a mesodermal lineage Naturewith therapeutic Reviews | Genetics potential and for generating pancreatic precursors as a key endodermal derivative that is crucial for clinical translation. Small molecules and growth factors that are used to direct cell fate are indicated below the arrows. Protocols for both lineages are rather short and require an enrichment step before transplantation. a | In the generation of cardiomyocytes from human embryonic stem cells (hESCs)4, a critical step is the induction of cardiac mesoderm in the presence of molecules that activate both bone morphogenetic protein (BMP) and Nodal– activin–transforming growth factor-β (TGFβ) signalling. Immature cardiomyocytes emerge by day 10 (d 10) of differentiation and are isolated by day 16. They are purified using Percoll, which is a physical separation technique. A critical step for the generation of cardiac cells is to show their ability to electrically couple with the host heart, as demonstrated using a genetic indicator of calcium signalling (GCaMP3). b | In the generation of hESC-derived pancreatic precursors30,38, induction of the endoderm is achieved in the presence of molecules that trigger WNT and Nodal–activin–TGFβ signalling pathways. Pancreatic endoderm is obtained by day 12 of differentiation, the stage at which cells can be either further matured in vitro or directly transplanted in vivo. Although in vivo maturation of CD142+ cells results in a large number of insulin-positive cells with appropriate glucose response, current techniques for in vitro maturation do not yield cells with proper glucose responsiveness. Cyc, cyclopamine; FACS, fluorescence-activated cell sorting; FBS, fetal bovine serum; RA, retinoic acid.

showed extensive remyelination in vivo and signifi- in vivo long-term survival and functional integration cantly extended lifespan in the transplanted mice. despite promising marker data in vitro. Here, we briefly However, the need for extensive in vitro differentiation discuss current progress in the development of hPSC- before transplantation (that is, 125 days) may hamper based cell therapies for cardiac repair and islet cell clinical translation. Furthermore, it may be advanta- replacement as two examples in which sufficient pro- geous to develop culture conditions that exclusively gress has been made to support clinical translation in yield oligodendrocytes, as the current grafts also con- the foreseeable future despite considerable remaining tain many astrocytes. Finally, it will be important to challenges (FIG. 2). graft these derived cells into specific locations of the adult CNS, where cell migration is likely to be more Cardiac cells. The derivation of cardiomyocytes — the limited, and into large animal models where the scale cells that make up the cardiac muscle — has been an of remyelination is more challenging. Several therapies important goal in hPSC research given the fact that are currently being developed using hPSC-derived cardiovascular disease represents the leading cause of oligodendroglial cells; these include treatment for rare death worldwide. There has been considerable progress genetic disorders, late-stage multiple sclerosis and in the past few years in developing defined protocols for patients with cancer who suffer from the long-term both cardiac specification and the derivation of func- side effects of brain irradiation (reviewed in REF. 27). tional cardiomyocytes. One of the two most successful strategies that are currently being developed to generate Deriving non-neural cell types cardiac cells at high efficiencies is a monolayer-based The therapeutic development for many cell lineages differentiation32 that is based on the sequential expo- outside the nervous system — such as hPSC-derived sure of differentiating hPSCs to activin A and BMP4. liver28, endothelial cells29 and pancreatic islets30, as well The other strategy is an embryoid body‑based differ- as engraftable haematopoietic stem cells31 — have hit entiation paradigm that follows a more complex series major ‘roadblocks’. These challenges include limited of defined patterning signals33.

NATURE REVIEWS | GENETICS VOLUME 15 | FEBRUARY 2014 | 85

Both strategies are currently under preclinical devel- is suitable for clinical use. This problem is particularly opment, and a recent paper has shown robust survival acute for cell types that develop their full functionality of hESC-derived cardiomyocytes and the ability of only at later stages of development, such as haematopoie these cells both to couple with host cells and to suppress tic stem cells with adult homing capacity, cardiac cells arrhythmias in injured hearts4 (FIG. 2a). Ongoing studies with adult-like physiology and pancreatic cells that show confirm robust survival of hESC-derived cardiomyo- normal regulation in response to glucose stimulation. cytes but report the induction of transient arrhythmias, Current differentiation strategies for hPSCs yield differ- which disappears completely over a period of weeks entiated cell types that most closely match embryonic to months. Immature cells transiently express features or fetal stages of development. Therefore, technologies of pacemaker cells that may result in a propensity to that enable the production of cell types with adult-like autonomously trigger cardiac contractions. The devel- functional properties need to be developed in the future. opment of methods to increase cell maturation, such as Harnessing the signalling pathways that control nor- through the use of tissue engineering approaches34, will mal development has developed into a powerful strategy be a major focus towards the clinical translation of this to direct hPSC fate. An obvious question is whether a technology in humans. better understanding of the developmental requirements will enable researchers to overcome the remaining road- Pancreatic islet cells. The development of a cell-based blocks in hPSC differentiation and to establish protocols therapy for type I diabetes has been at the forefront of for any of the cell types in our body. In some instances, efforts in pluripotent stem cell research. Similarly to such as in the case of skeletal muscle precursors39, early Parkinson’s disease and Huntington’s disease, there has hPSC-based protocols were not robust enough to yield been a history of fetal cell-grafting approaches and, in large numbers of engraftable myoblasts. However, more the case of islet cell replacement, grafting of cadaver efficient engraftment, including functional integra- material (for example, the Edmonton protocol). A spe- tion into the muscle of dystrophic mice, was observed cific problem in type I diabetes is the need to address the in hESC-derived cells with an inducible expression of underlying autoimmune disorder, in addition to replac- paired box protein PAX‑7 (REF. 40). It will be interesting ing the specific cells that are lost owing to the disease. In to see whether refined patterning strategies will enable the past few years, excellent protocols have been devel- the derivation of differentiated cells, including skeletal oped for the induction of early endodermal and pancre- muscle, using purely extrinsic signals40 or whether there atic lineages from hPSCs35–37. The main challenge has remains a need for either transient or stable genetic been the derivation of engraftable cells that are glucose modification in the specification of certain cell lineages. responsive, that show high levels of insulin production Examples of new developments in directed differentiation and that do not co‑express other hormones. A strategy are discussed below (FIG. 3). to bypass the limited functionality of hPSC-derived, insulin-producing β‑cells is the transplantation of pan- Small molecules. The use of small molecules was an creatic precursor cells that are subsequently matured important tool in improving hPSC-based directed in vivo38 (FIG. 2b). Although in vivo maturation seems to differentiation protocols (reviewed in REF. 41). Small facilitate the derivation of functional β‑cells, the use of molecules are typically less costly than recombinant proliferating precursors might result in additional risks protein factors and show higher potency and scal- for clinical translation. Therefore, current translational ability42; they also enable the direct manipulation of efforts include the development of an encapsulation intracellular pathways. Interestingly, off-target effects technology to address concerns about cell overgrowth. have not caused substantial problems in directed dif- Another valuable step in the development of a clinically ferentiation paradigms. Another surprising feature of meaningful differentiation protocol is the use of surface small-molecule-based studies is that, in several cases, markers such as CD142 to enrich for pancreatic precur- the inhibition rather than the activation of signalling sor cells that have the highest capacity for in vivo islet cell pathways is crucial for fate specification. In the case of production30. These strategies may enable the current neural induction, the combined inhibition of BMP and protocols to move towards early-stage clinical studies. Nodal–activin–TGFβ signalling (that is, dSMADi11) has However, the development of a long-term, competitive emerged as a powerful platform for generating specific hPSC-based islet-cell therapy for the routine treatment neural lineages. The key advantage of this approach is of type I diabetes is likely to require new technology the highly synchronized and efficient nature of neural to obtain hPSC-derived cells that are fully scalable induction that enables precise temporal patterning of and functional without the need for extensive in vivo cells1. A similar approach has been adopted for the proliferation and maturation. specification of endodermal derivatives, such as cells of the lung and thyroid lineages43. To push the limits Considerations for cell differentiation of this approach, we have recently described a rapid Choice of cell source and developmental consideration. induction protocol that yields hPSC-derived neurons Pluripotent stem cells represent a particularly attrac- within ten days of differentiation using a combina- tive cell source given their scalability and versatility. tion of five inhibitors44. Medically important neurons, However, for most hPSC-derived lineages, a major chal- such as pain-sensing nociceptors, were thus derived lenge that remains is to harness the broad differentiation at an unprecedented speed by targeting BMP, Nodal– potential of these cells to generate a cell product that activin–TGFβ, fibroblast growth factor (FGF), Notch and

86 | FEBRUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

Tools to direct hPSC diﬀerentiation Development in a dish Quality control of diﬀerentiated progeny

Small molecules and recombinant proteins Germ-layer speciﬁcation Cytochemistry and ultrastructure NH H3 C N N hPSC H N N N N H CI CI CN D e.g. immunocytochemistry (markers of fate, Inhibition or activation: WNT, SHH, BMP, TGFβ, Ect Me En regional identity and maturation) and Notch, FGF, IGF, EGF and PDGF electron microscopy A P V Matrices and self-patterning Genetic and epigenetic identity markers Anterioposterior and dorsoventral identity

Terminal fate specifcation Organoids, co-culture, tissue induction, and isolation e.g. proﬁling mRNA, protein, DNA and histone integrins, matrigel and hydrogels methylation and bulk versus single-cell studies

Transcriptional programming Biochemical and physiological assays

LMX1A NR4A2 FOXA2 ASCL1 FACS, MACS, survival and maturation towards adult-like function? Towards terminal fate, somatic stem cell and e.g. Ca2+-imaging, electrophysiology, hormone or in vivo reprogramming In vitro maturation neurotransmitter release and optogentic tools

Figure 3 | Directing fate and validating identity of hPSC-derived lineages. The key tools that are used to direct Nature Reviews | Genetics the differentiation of human pluripotent stem cells (hPSCs) to somatic cell types are shown. The use of small- molecule-based approaches has been particularly useful in generating cells with potential therapeutic relevance. Most strategies for directed cell differentiation are based on recreating aspects of normal development in vitro. One key developmental decision during the differentiation of hPSCs is the initial specification towards one of the three germ layer derivatives (that is, lineages of the ectoderm (ect), endoderm (en) or mesoderm (me)). A second crucial decision for hPSCs is to acquire specific anterioposterior (A–P) and dorsoventral (D–V) patterning fates. By recreating those specific signalling conditions, it is possible to generate hPSC-derived lineages that correspond to the cells that originate in the different regions of the developing embryo. However, despite such progress, a limited understanding of human development and its protracted timeframes remain important ‘roadblocks’ in the field. Assessment of the authenticity of cell fate in the differentiated progeny is a crucial element of any directed differentiation strategy. In addition to traditional methods such as cytochemistry and gene expression studies, there is an increased need for robust in vitro functional assays. ASCL1, achaete-scute homologue 1; BMP, bone morphogenetic protein; EGF, epidermal growth factor; FACS, fluorescence-activated cell sorting; FGF, fibroblast growth factor; FOXA2, forkhead box protein A2; IGF, insulin-like growth factor; LMX1A, LIM homeobox transcription factor 1‑α; MACS, magnetic activated cell sorting; NR4A2, nuclear receptor subfamily 4 group A member 2; PDGF, platelet-derived growth factor; SHH, sonic hedgehog; TGFβ, transforming growth factor-β.

WNT signalling pathways. Combined small-molecule Self-organization and differentiation in three dimen- approaches may be useful for other hPSC-derived sions. The establishment of proper three-dimensional lineages, such as pancreatic precursors35. The ability cytoarchitecture is a crucial feature for several cell to shorten the timing of human cell-fate specification types. For example, although the first clinical trials using in vitro, as shown for the rapid induction of human hPSC-derived retinal pigment epithelial (RPE) cells for nociceptors44, may be a useful strategy to simplify the treating macular degeneration were based on injecting induction of both late-born and late-maturing cell dissociated cells47, there are extensive efforts to gener- types, such as oligodendrocytes and cortical interneu- ate sheets of RPE cells to ensure proper organization of rons. The effect of small-molecule approaches was these cells following transplantation into the patient’s particularly marked for improving direct neural induc- eye. The formation of three-dimensional aggregates has tion protocols. However, small molecules are widely been a feature of several recent directed differentiation used in various neural induction paradigms, including protocols, such as the derivation of pituitary cells from feeder-based45 and embryoid body-based46,47 neural mouse embryonic stem cells, the generation of human differentiation protocols for hESCs or iPSCs46. or mouse optic cup-like structures and the derivation

NATURE REVIEWS | GENETICS VOLUME 15 | FEBRUARY 2014 | 87

of cortical and cerebral organoid48 cultures, as well as antibody-based isolation strategy is the use of surface intestinal crypt structures, from hESCs (reviewed in marker antibody screens, whereby 242–370 commer- REF. 49). Under those conditions, the main strategy is to cially available antibodies are screened against cells in enable differentiating hPSCs to reveal their self-organizing a specific stage of differentiation. This strategy was use- properties without the requirement for any specific ful for developing enrichment strategies for neural stem extrinsic cues. Parameters that do affect the differen- cells and neurons61, as well as for various non-neural tiation process in a self-organizing paradigm include lineages such as hESC-derived cardiomyocytes62. initial cell density and overall medium composition. Some of those studies also rely on the use of specific Defining cell fate and function extracellular matrix components, such as hPSC-derived Novel assays to characterize cell fate in vitro. The iden- cerebral organoids that require embedding early neu- tification of defined cell types and stages in vitro is an ral rosettes in a matrigel plug before the formation of important component of any directed differentiation organoid structures. Directing differentiation of hPSCs strategy. Clearly, it is important to define the expression by manipulating extracellular matrix components has of cell type-specific markers at both the gene and pro- developed into a strategy for various hPSC-derived lin- tein levels. Similarly, it is vital to confirm the absence eages, such as vascular cells50, bone51 and cartilage52, and of markers that should not be expressed in a given cell such approach is probably applicable to many other line- type. Flow-based quantification of marker expression ages as reviewed in REF. 53. However, the relative require- is particularly useful, as it circumvents the challenges ments for three-dimensional organization are variable. that are associated with image analyses in cultures plates For example, a recent study using hESC-derived cardio- that contain a mixture of cell clusters and with individual myocytes reported evidence for functional integration cells that are difficult to quantify. However, beyond sim- and coupling of grafted cardiomyocytes with host cells ple candidate marker expression, unbiased genome-wide in the guinea pig heart without the requirement of cell analyses for transcription, such as microarray and RNA assembly before grafting4. Similarly, fetal midbrain- sequencing platforms, are increasingly being used as a grafting studies in models of Parkinson’s disease, which standard in the field. There is also an increasing effort is considered to be the ‘gold standard’ for future hPSC- both to include global assays that define the chromatin grafting paradigms, did not show any benefit in grafting state of differentiated hPSC progeny, as recently shown tissue pieces with intact cytoarchitecture compared to for hPSC-derived cardiac precursors63, and to define cell suspension grafts54. One example in which three- the enhancer landscape in hPSC-derived neural crest dimensional organization may have partially overcome precursors64. As directed differentiation of hPSCs relies a functional bottleneck in hPSC differentiation is the on developmental signalling events that control normal establishment of engraftable liver bud-like structures. In embryonic development, the demonstration that hPSC- this example, co‑culture with endothelial and mesenchy- derived cultures transit through appropriate develop- mal cells resulted in structures that showed improved mental intermediate stages is also important in assessing functionality in producing key liver enzymes55. the authenticity of the final differentiated cell fate. Self-organizing Pertaining to an intrinsic programme in pluripotent Cell purification. There has been extensive progress in Assays to assess in vivo survival and function. Early stem cell-derived lineages developing genetic reporter lines to optimize directed transplantation studies were mostly limited to mea that enables cells in vitro to differentiation and to prospectively purify defined cell suring cell survival by histology and by correlating assemble into tissue-like types for downstream applications. Although early those results with behavioural outcome. However, new and organoid structures. studies were commonly based on either small plasmid- imaging and functional tools aim to change the field Transcription activator-like based reporter constructs or transgenesis using bacte- by enabling real-time monitoring of graft survival, effector nucleases rial artificial chromosomes (reviewed in REF. 56), more function and cellular integration. Advances in positron (TALENs). Fusions of truncated recent studies have focused on the use of reporter lines emission tomography (PET), optical imaging and mag- TALEs to a nonspecific DNA-cleavage domain of the that are established by gene targeting given the increas- netic resonance imaging (MRI) have steadily pushed FokI endonuclease. Each TALE ing efficiency at which homologous recombination the boundaries of spatial resolution, especially with the contains an amino terminus, a can be achieved in hPSCs using technologies that are use of nanoparticles, thus allowing the development of custom-designed DNA-binding based on zinc-finger nucleases, transcription activator-like accurate spatiotemporal maps of grafts and even at a domain and a carboxyl effector nucleases (TALENs) or clustered regularly inter- single-cell level in some cases65. Some of those tech- terminus with the activation domain being removed. spaced short palindromic repeats (CRISPRs) (reviewed nologies have not fully reached the clinic yet, but their in REF. 57). Such studies have been useful in defining preclinical development is evolving at a rapid pace66. Positron emission the appropriate stage of midbrain dopamine neurons Furthermore, bioluminescent imaging and magnetic tomography for grafting by comparing three genetically defined labelling of grafted cells have been successfully used (PET). An imaging technique that detects the emission of stages from dividing precursors to differentiated neu- to track human neural, endothelial and skeletal muscle 58 39 positrons from the brain after rons . Other recent examples include the isolation of precursors in vivo. Strategies that are currently under a small amount of radioactive homeobox protein NKX‑2.1–GFP+ cells for subsequent development for use in hPSC-derived lineages include isotopes have been injected derivation of highly enriched populations of cortical genetic reporters that allow induced-fate mapping and into the blood stream; it is interneurons59,60. However, for translational applica- in vivo clonal analyses, which are routinely used in used to quantitatively measure 67 metabolic, biochemical and tions it may be beneficial to avoid the genetic modifi- mice . Genetic tracking has already been used clini- functional activity in living cation of cells. A powerful technology to facilitate the cally during experimental brain tumour therapies with tissues. transition from a genetic-based isolation strategy to an vectors that allow both targeting of the tumour, such as

88 | FEBRUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

ganciclovir-induced activation of herpes simplex virus methylation80. However, the presence of genetic and thymidine kinase (HSV‑tk), and imaging using PET epigenetic aberrations is not limited to human iPSCs modalities68. Furthermore, in cell therapies following and to cellular reprogramming but is a concern for any introduction of autologous genetically modified haema- cell-based therapeutic product. topoietic stem cells, the integration site of the genetic Recent advances in reprogramming techniques obvi- vector itself is a useful reporter to study the distribution ate the need for integrating vectors81–83, thereby eliminat- and the persistence of grafted cells. ing the risk of reactivation of oncogenic reprogramming Traditional strategies for assessing graft function factors. Another concern that is specific to the use include electrophysiology, systemic measurements of of iPSC-derived cells is whether reprogrammed cells hormone release and the use of in vivo microdialysis retain an epigenetic ‘memory’. There is clear evidence to determine neurotransmitter levels69. Recent tools of that both mouse and human iPSCs retain, at least tran- increasing importance include optogenetic reporters, siently, an epigenetic marker profile that is partly related which allow manipulation of neuronal activity in grafted to the donor cell of origin and that can affect subsequent cells through extrinsic light pulses. These reporters have differentiation results84–87. Future studies will have to been used to show connectivity of hPSC-derived neu- address whether the simple maintenance of iPSCs at the rons70. Similarly, the use of genetically encoded GCaMP3 pluripotent stage can fully eliminate epigenetic memory, calcium sensors was valuable in monitoring functional whether additional pharmacological or genetic interven- coupling of hPSC-derived cardiomyocytes in the host tions are required for full reset and whether any transla- heart4. Genetic ablation studies are also important for tional application of iPSC-derived lineages in cell therapy determining the mechanisms that underlie graft func- will require a careful analysis of the effect of donor cell tion, such as by distinguishing short-term trophic effects origin on cell differentiation and function. from effects that are due to long-term functional engraft- Some of the biggest practical hurdles for using ment. Suicide genes, such as those encoding HSV‑tk patient-specific iPSCs are regulatory requirements, as and diphtheria toxin, have been successfully used to well as the cost and timeframes that are associated with demonstrate a specific role for grafted neural stem cells approving unique iPSC-based products for each indi- in spinal cord injury models71. These genes have also vidual. Given the current technologies and the exist- been proposed as clinical fail-safe mechanism in case of ing regulatory framework, the costs that are associated overproliferation of grafted cells. More recent examples with such a personalized treatment approach would be include the use of induced caspase‑9, which is a trigger prohibitive, which makes it unlikely for such approach that is suitable for clinical use72 and for eliminating both to enter routine, reimbursable medical practice. Both proliferating and postmitotic cells. Genetic engineering technological and regulatory changes will be required is likely to have important roles beyond safety switches to to overcome these important hurdles in the future. keep ‘wayward cells’ in check. Genetically modified cells Nonetheless, researchers in Japan are currently seek- can deliver therapeutically relevant proteins; for exam- ing regulatory approval for the first of such study using ple, neural precursors that are genetically engineered patient-specific iPSC-derived RPE cells for the treatment to express glial cell line-derived neurotrophic factor of macular degeneration88. The recent breakthrough in (GDNF) are currently being developed as a therapeutic establishing patient-matched hPSCs through nuclear strategy for treating patients with amyotrophic lateral transfer using a Dolly-like cloning approach89 puts sclerosis73. Other important applications of genetically another potential cell source back into play for the deve engineered stem cells include repair of a disease-causing lopment of patient-matched therapies through nuclear gene, such as the replacement of β‑globin in thalas- transfer of hESCs. semia74, as well as the modulation of immunogenicity Directed conversion of somatic cells through the of the grafted cells or of the ability to express factors introduction of fate-specific transcription factors has that could promote graft integration. A recent example been reported for several fates as reviewed recently8,90. for modulating graft integration and axonal outgrowth However, the translational potential of this approach, is the use of hESC-derived dopamine neurons that including the ability for long-term functional engraft- are modified to express PST (also known as SIA8D), ment, remains to be determined. The scalability and the which is the enzyme required to regulate polysialylated establishment of a good manufacturing practice (GMP)- neuronal cell adhesion molecules at the cell surface75. compliant cell therapy platform represent additional major hurdles towards clinical translation91. Another Autologous cell sources. Patient-specific iPSCs are attrac- future research direction involves in vivo reprogram- tive as an essentially autologous source that should obvi- ming with a goal of targeting the body’s own cells. These ate the need for immunosuppression in the recipient, efforts are currently limited to early preclinical studies,

Good manufacturing although this issue remains somewhat controversial such as the conversion of different neuronal sub- 76–78 92,93 94 practice on the basis of studies in mouse iPSCs . Full immu- types , the conversion of astrocytes into neurons (GMP). A set of standardized nocompatibility is difficult to assess for human cells or neuroblasts95, the in vivo conversion of exocrine to production and testing because of the lack of an experimental autologous graft- endocrine pancreatic cells96 and the in vivo conver- conditions that are required for ing paradigm. A potential concern for using human sion of cardiac fibroblasts into cardiomyocytes97. All developing a clinical-grade cell product and for obtaining iPSCs is the acquisition of genetic and epigenetic altera- of these applications hold substantial promise for the regulatory approval for trials tions during reprogramming, including mutations in future, although none of them seems to be close to in human subjects. protein-coding regions79 and evidence of aberrant DNA clinical translation.

NATURE REVIEWS | GENETICS VOLUME 15 | FEBRUARY 2014 | 89

Preclinical proof of concept GMP production and scaling

Clinical trial design • Functional readout • GMP-compliant hPSCs and • Relevant animal model diﬀerentiation protocols • Mechanism of action • Master and working banks • Patient selection (that is, • Enhancement strategies? (small, intermediate and subclassify patient groups) clinical sizes) • Deﬁne standardized readout

• Survival and function Validation and toxicity Phase I/II clinical trial • Imaging, histology and behaviour • Immunology • Validation of preclinical data • Safety (procedure, cell and dosage) • HLA-matched versus • Large animal models • Flexible design and ‘futility’ trial non-HLA-matched hPSCs • Dose escalation • Surrogate markers of eﬃcacy • Safety and side eﬀects • Long-term toxicity • Tumour risk, inappropriate fates and overdosing (genetic safety switches?) Phase III/IV clinical trial

• Double-blind trials of clinical eﬃcacy • Compare with state-of-the-art treatments • Costs and long-term follow-up

Figure 4 | A roadmap towards clinical translation. To move human pluripotent stem cell (hPSC)-derivedNature Reviews cells | Genetics from the bench to clinical studies, some key steps and challenges need to be addressed. Following a preclinical proof of concept, it is important to validate robustness in relevant disease models and to develop a cell-manufacturing strategy that is suitable for clinical translation. Important steps include producing cells at a sufficient scale (that is, creating a cell bank) and using fully standardized protocols under current good manufacturing practice (GMP)-compliant conditions. The resulting bank of the candidate therapeutic cell product needs to be revalidated for safety and efficacy before it can be considered as a candidate clinical product for early safety studies, and eventually efficacy studies, in human patients. HLA, human leukocyte antigen.

Next steps towards translation manner and, crucially, the selection of appropriate dis- In the past few years, there has been a strong drive ease targets. Substantial progress has been made in the towards translating hESC research into the clinic. area of scalable production of hESC progeny using xeno- The first of such attempt was the use of hESC-derived free reagents, defined culture media and robust cryo- oligodendrocyte precursors for treating spinal cord preservation techniques, all of which were carried out injury with the goal of remyelinating denuded axons under GMP-compliant conditions99. at the site of injury. The trial, which was sponsored by Diseases that are caused by the selective loss of Geron, has pioneered the application of hESCs and specific cell populations have been at the forefront illustrated the challenges for passing the regulatory of clinical developments. However, precise patient hurdles in this new class of therapy. Although the spi- selection often remains a major concern. In the case nal cord trial has been abandoned owing to financial of Parkinson’s disease40, incomplete knowledge of and strategic considerations at Geron, the lack of major the pathophysiology and of the potential clinical and adverse effects in the four patients treated offers pre- molecular subsets of the disease is compounded by liminary safety data for the approach98. The second poorly representative animal models. Thus, improved application for a hESC-derived product was the use of phenotyping of patients using physiological, genetic RPE cells for treating eye disorders such as Stargardt and imaging approaches is crucial for selecting appro- disease — a juvenile form of macular degeneration — priate subjects for cell therapy. Furthermore, any cell- and age-related macular degeneration. These ongoing based approach has to be competitive with existing studies have reported survival of the cells in at least one therapies, such as novel pharmacological agents, or of the patients who was initially grafted and associated with alternative surgical interventions such as deep improvement in visual function47. brain stimulation100. Beyond these few initial attempts, realizing the thera Grafting techniques will benefit from remarkable peutic potential of stem cells for clinical applications technical innovations in the surgical field, such as the remains a central goal for the scientific community and development of robotic platforms and minimally inva- the public. Two of the key elements for any successful sive delivery systems. For example, for intracranial translational application (FIG. 4) are the ability to pro- injections, recent advances include the development duce hPSC derivatives in a scalable and GMP-compliant of surgical technologies that provide real-time MRI

90 | FEBRUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics

guidance and visualization of the brain target101. The potential risks of extended biological activity, induction ability to monitor grafted cells could be coupled with of immunogenicity and the involvement of relatively non-invasive methods to control graft behaviour, such invasive procedures and devices for delivery. as the inducible safety switches discussed above. Finally, methods are under development to enhance in vivo cell Conclusion migration and graft integration. Implementing hPSC- Stem cell research has moved from an era of optimizing based approaches in regenerative medicine will require hPSC isolation to ongoing or imminent early-stage clini- multidisciplinary teams of clinicians and scientists with cal trials. A major challenge that remains is to define a expertise in directed differentiation, GMP produc- more comprehensive set of human disorders that may tion, large animal studies, tissue engineering and trial benefit from hPSC-based approaches. Furthermore, design, as well as ethicists and patient advocates. In directed differentiation will have to include methods to addition, navigating the complex web of federal regula- control cell maturation, in addition to cell-fate specifica- tions requires specific expertise, especially in view of the tion, in order to access a full range of cell types and stages pioneer status of these approaches. In fact, the Center for applications in regenerative medicine. Finally, safety for Biologics Evaluation and Research of the US Food concerns remain an important issue despite the promise and Drug Administration (FDA) has recently published of preclinical studies. After 15 years of hESC research a guidance102 entitled “Considerations for the design of and six years since the isolation of human iPSCs, we early-phase clinical trials of cellular and gene therapy are on the verge of defining the clinical use of hPSCs in products”, in which the FDA expressed concerns over the human disease.

1. Kriks, S. et al. Dopamine neurons derived from human 18. Kirkeby, A. et al. Generation of regionally specified 33. Yang, L. et al. Human cardiovascular progenitor cells ES cells efficiently engraft in animal models of neural progenitors and functional neurons from develop from a KDR+ embryonic-stem-cell-derived Parkinson’s disease. Nature 480, 547–551 (2011). human embryonic stem cells under defined conditions. population. Nature 453, 524–528 (2008). 2. Ma, L. et al. Human embryonic stem cell-derived Cell Rep. 1, 703–714 (2012). 34. Nunes, S. S. et al. Biowire: a platform for maturation GABA neurons correct locomotion deficits in quinolinic 19. Xi, J. et al. Specification of midbrain dopamine of human pluripotent stem cell-derived cardiomyocytes. acid-lesioned mice. Cell Stem Cell 10, 455–464 (2012). neurons from primate pluripotent stem cells. Stem Nature Methods 10, 781–787 (2013). 3. Wang, S. et al. Human iPSC-derived oligodendrocyte Cells 30, 1655–1663 (2012). 35. Chen, S. et al. A small molecule that directs progenitor cells can myelinate and rescue a mouse 20. MacDonald, M. E. et al. A novel gene containing a differentiation of human ESCs into the pancreatic model of congenital hypomyelination. Cell Stem Cell trinucleotide repeat that is expanded and unstable lineage. Nature Chem. Biol. 5, 258–265 (2009). 12, 252–264 (2013). on Huntington’s disease chromosomes. Cell 72, 36. D’Amour, K. A. et al. Production of pancreatic 4. Shiba, Y. et al. Human ES‑cell-derived cardiomyocytes 971–983 (1993). hormone-expressing endocrine cells from human electrically couple and suppress arrhythmias in injured 21. Perrier, A. & Peschanski, M. How can human embryonic stem cells. Nature Biotech. 24, hearts. Nature 489, 322–325 (2012). pluripotent stem cells help decipher and cure 1392–1401 (2006). 5. Bellin, M., Marchetto, M. C., Gage, F. H. & Huntington’s disease? Cell Stem Cell 11, 153–161 37. Nostro, M. C. et al. Stage-specific signaling through Mummery, C. L. Induced pluripotent stem cells: (2012). TGFβ family members and WNT regulates patterning the new patient? Nature Rev. Mol. Cell Biol. 13, 22. Aubry, L. et al. Striatal progenitors derived from and pancreatic specification of human pluripotent 713–726 (2012). human ES cells mature into DARPP32 neurons in vitro stem cells. Development 138, 861–871 (2011). 6. Zhang, S. C., Wernig, M., Duncan, I. D., Brustle, O. & and in quinolinic acid-lesioned rats. Proc. Natl Acad. 38. Kroon, E. et al. Pancreatic endoderm derived from Thomson, J. A. In vitro differentiation of Sci. USA 105, 16707–16712 (2008). human embryonic stem cells generates glucose- transplantable neural precursors from human 23. Carri, A. D. et al. Developmentally coordinated responsive insulin-secreting cells in vivo. Nature embryonic stem cells. Nature Biotech. 19, extrinsic signals drive human pluripotent stem cell Biotech. 26, 443–452 (2008). 1129–1133 (2001). differentiation toward authentic DARPP‑32+ medium- 39. Barberi, T. et al. Derivation of engraftable skeletal 7. Reubinoff, B. E. et al. Neural progenitors from sized spiny neurons. Development 140, 301–312 myoblasts from human embryonic stem cells. Nature human embryonic stem cells. Nature Biotech. 19, (2013). Med. 13, 642–648 (2007). 1134–1140 (2001). 24. Nistor, G. I., Totoiu, M. O., Haque, N., Carpenter, M. K. 40. Darabi, R. et al. Human ES- and iPS-derived myogenic 8. Ladewig, J., Koch, P. & Brustle, O. Leveling & Keirstead, H. S. Human embryonic stem cells progenitors restore DYSTROPHIN and improve Waddington: the emergence of direct programming differentiate into oligodendrocytes in high purity and contractility upon transplantation in dystrophic mice. and the loss of cell fate hierarchies. Nature Rev. Mol. myelinate after spinal cord transplantation. Glia 49, Cell Stem Cell 10, 610–619 (2012). Cell Biol. 14, 225–236 (2013). 385–396 (2005). 41. Li, W., Li, K., Wei, W. & Ding, S. Chemical approaches 9. Watanabe, K. et al. A ROCK inhibitor permits survival 25. Alper, J. Geron gets green light for human trial of ES to stem cell biology and therapeutics. Cell Stem Cell of dissociated human embryonic stem cells. Nature cell-derived product. Nature Biotech. 27, 213–214 13, 270–283 (2013). Biotech. 25, 681–686 (2007). (2009). 42. Li, W. & Ding, S. Small molecules that modulate 10. Perrier, A. L. et al. Derivation of midbrain dopamine 26. Hu, B. Y., Du, Z. W., Li, X. J., Ayala, M. & Zhang, S. C. embryonic stem cell fate and somatic cell neurons from human embryonic stem cells. Proc. Natl Human oligodendrocytes from embryonic stem cells: reprogramming. Trends Pharmacol. Sci. 31, 36–45 Acad. Sci. USA 101, 12543–12548 (2004). conserved SHH signaling networks and divergent FGF (2010). 11. Chambers, S. M. et al. Highly efficient neural effects. Development 136, 1443–1452 (2009). 43. Longmire, T. A. et al. Efficient derivation of purified conversion of human ES and iPS cells by dual 27. Goldman, S. A., Nedergaard, M. & Windrem, M. S. lung and thyroid progenitors from embryonic stem inhibition of SMAD signaling. Nature Biotech. 27, Glial progenitor cell-based treatment and modeling cells. Cell Stem Cell 10, 398–411 (2012). 275–280 (2009). of neurological disease. Science 338, 491–495 44. Chambers, S. M. et al. Combined small-molecule 12. Lindvall, O. Dopaminergic neurons for Parkinson’s (2012). inhibition accelerates developmental timing and therapy. Nature Biotech. 30, 56–58 (2012). 28. Touboul, T. et al. Generation of functional hepatocytes converts human pluripotent stem cells into 13. Olanow, C. W. et al. A double-blind controlled trial of from human embryonic stem cells under chemically nociceptors. Nature Biotech. 30, 715–720 (2012). bilateral fetal nigral transplantation in Parkinson’s defined conditions that recapitulate liver development. 45. Morizane, A., Doi, D., Kikuchi, T., Nishimura, K. & disease. Ann. Neurol. 54, 403–414 (2003). Hepatology 51, 1754–1765 (2010). Takahashi, J. Small-molecule inhibitors of bone 14. Freed, C. R. et al. Transplantation of embryonic 29. James, D. et al. Expansion and maintenance of human morphogenic protein and activin/nodal signals dopamine neurons for severe Parkinson’s disease. embryonic stem cell-derived endothelial cells by TGFβ promote highly efficient neural induction from human N. Engl. J. Med. 344, 710–719 (2001). inhibition is Id1 dependent. Nature Biotech. 28, pluripotent stem cells. J. Neurosci. Res. 89, 117–126 15. Barker, R. A., Barrett, J., Mason, S. L. & Bjorklund, A. 161–166 (2010). (2011). Fetal dopaminergic transplantation trials and the 30. Kelly, O. G. et al. Cell-surface markers for the isolation 46. Boulting, G. L. et al. A functionally characterized test future of neural grafting in Parkinson’s disease. Lancet of pancreatic cell types derived from human embryonic set of human induced pluripotent stem cells. Nature Neurol. 12, 84–91 (2013). stem cells. Nature Biotech. 29, 750–756 (2011). Biotech. 29, 279–286 (2011). 16. Yan, Y. et al. Directed differentiation of dopaminergic 31. Ledran, M. H. et al. Efficient hematopoietic 47. Schwartz, S. D. et al. Embryonic stem cell trials for neuronal subtypes from human embryonic stem cells. differentiation of human embryonic stem cells on macular degeneration: a preliminary report. Lancet Stem Cells 23, 781–790 (2005). stromal cells derived from hematopoietic niches. Cell 379, 713–720 (2012). 17. Sonntag, K. C. et al. Enhanced yield of neuroepithelial Stem Cell 3, 85–98 (2008). 48. Lancaster, M. A. et al. Cerebral organoids model precursors and midbrain-like dopaminergic neurons 32. Laflamme, M. A. et al. Cardiomyocytes derived from human brain development and microcephaly. Nature from human embryonic stem cells using the bone human embryonic stem cells in pro-survival factors 501, 373–379 (2013). morphogenic protein antagonist noggin. Stem Cells enhance function of infarcted rat hearts. Nature 49. Sasai, Y. Cytosystems dynamics in self-organization of 25, 411–418 (2007). Biotech. 25, 1015–1024 (2007). tissue architecture. Nature 493, 318–326 (2013).

NATURE REVIEWS | GENETICS VOLUME 15 | FEBRUARY 2014 | 91

50. Kusuma, S. et al. Self-organized vascular networks 70. Weick, J. P., Liu, Y. & Zhang, S. C. Human embryonic 89. Tachibana, M. et al. Human embryonic stem cells from human pluripotent stem cells in a synthetic stem cell-derived neurons adopt and regulate the derived by somatic cell nuclear transfer. Cell 153, matrix. Proc. Natl Acad. Sci. USA 110, 12601–12606 activity of an established neural network. Proc. Natl 1228–1238 (2013). (2013). Acad. Sci. USA 108, 20189–20194 (2011). 90. Vierbuchen, T. & Wernig, M. Molecular roadblocks for 51. Marolt, D. et al. Engineering bone tissue from human 71. Cummings, B. J. et al. Human neural stem cells cellular reprogramming. Mol. Cell 47, 827–838 (2012). embryonic stem cells. Proc. Natl Acad. Sci. USA 109, differentiate and promote locomotor recovery in spinal 91. Chambers, S. M. & Studer, L. Cell fate plug and play: 8705–8709 (2012). cord-injured mice. Proc. Natl Acad. Sci. USA 102, direct reprogramming and induced pluripotency. 52. Diekman, B. O. et al. Cartilage tissue engineering 14069–14074 (2005). Cell 145, 827–830 (2011). using differentiated and purified induced pluripotent 72. Di Stasi, A. et al. Inducible apoptosis as a safety switch 92. Rouaux, C. & Arlotta, P. Direct lineage reprogramming stem cells. Proc. Natl Acad. Sci. USA 109, for adoptive cell therapy. New Engl. J. Med. 365, of post-mitotic callosal neurons into corticofugal 19172–19177 (2012). 1673–1683 (2011). neurons in vivo. Nature Cell Biol. 15, 214–221 (2013). 53. Kraehenbuehl, T. P., Langer, R. & Ferreira, L. S. 73. Suzuki, M. & Svendsen, C. N. Combining growth 93. De la Rossa, A. et al. In vivo reprogramming of circuit Three-dimensional biomaterials for the study of factor and stem cell therapy for amyotrophic lateral connectivity in postmitotic neocortical neurons. human pluripotent stem cells. Nature Methods 8, sclerosis. Trends Neurosci. 31, 192–198 (2008). Nature Neurosci. 16, 193–200 (2013). 731–736 (2011). 74. Papapetrou, E. P. et al. Genomic safe harbors permit 94. Torper, O. et al. Generation of induced neurons via 54. Redmond, D. E. Jr., Vinuela, A., Kordower, J. H. & high β-globin transgene expression in thalassemia direct conversion in vivo. Proc. Natl Acad. Sci. USA Isacson, O. Influence of cell preparation and target induced pluripotent stem cells. Nature Biotech. 29, 110, 7038–7043 (2013). location on the behavioral recovery after striatal 73–78 (2011). 95. Niu, W. et al. In vivo reprogramming of astrocytes to transplantation of fetal dopaminergic neurons in a 75. Battista, D., Ganat, Y., El Maarouf, A., Studer, L. & neuroblasts in the adult brain. Nature Cell Biol. 15, primate model of Parkinson’s disease. Neurobiol. Dis. Rutishauser, U. Enhancement of polysialic acid 1164–1175 (2013). 29, 103–116 (2008). expression improves function of embryonic stem- 96. Zhou, Q., Brown, J., Kanarek, A., Rajagopal, J. & 55. Takebe, T. et al. Vascularized and functional human derived dopamine neuron grafts in Parkinsonian mice. Melton, D. A. In vivo reprogramming of adult liver from an iPSC-derived organ bud transplant. Stem Cells Transl Med. http://dx.doi.org/10.5966/ pancreatic exocrine cells to β-cells. Nature 455, Nature 499, 481–484 (2013). sctm.2013-0084 (2013). 627–632 (2008). 56. Giudice, A. & Trounson, A. Genetic modification of 76. Araki, R. et al. Negligible immunogenicity of terminally 97. Qian, L. et al. In vivo reprogramming of murine human embryonic stem cells for derivation of target differentiated cells derived from induced pluripotent or cardiac fibroblasts into induced cardiomyocytes. cells. Cell Stem Cell 2, 422–433 (2008). embryonic stem cells. Nature 494, 100–104 (2013). Nature 485, 593–598 (2012). 57. Gaj, T., Gersbach, C. A. & Barbas, C. F. 3rd. ZFN, 77. Zhao, T., Zhang, Z. N., Rong, Z. & Xu, Y. 98. Frantz, S. Embryonic stem cell pioneer Geron exits TALEN, and CRISPR/Cas-based methods for genome Immunogenicity of induced pluripotent stem cells. field, cuts losses. Nature Biotech. 30, 12–13 (2012). engineering. Trends Biotechnol. 31, 397–405 (2013). Nature 474, 212–215 (2011). 99. Ausubel, L. J., Lopez, P. M. & Couture, L. A. 58. Ganat, Y. M. et al. Identification of embryonic stem 78. Guha, P., Morgan, J. W., Mostoslavsky, G., GMP scale‑up and banking of pluripotent stem cells cell-derived midbrain dopaminergic neurons for Rodrigues, N. P. & Boyd, A. S. Lack of immune for cellular therapy applications. Methods Mol. Biol. engraftment. J. Clin. Invest. 122, 2928–2939 response to differentiated cells derived from syngeneic 767, 147–159 (2011). (2012). induced pluripotent stem cells. Cell Stem Cell 12, 100. Okun, M. S. Deep-brain stimulation for Parkinson’s 59. Nicholas, C. R. et al. Functional maturation of hPSC- 407–412 (2013). disease. New Engl. J. Med. 367, 1529–1538 (2012). derived forebrain interneurons requires an extended 79. Gore, A. et al. Somatic coding mutations in human 101. Larson, P. S. et al. An optimized system for timeline and mimics human neural development. Cell induced pluripotent stem cells. Nature 471, 63–67 interventional magnetic resonance imaging-guided Stem Cell 12, 573–586 (2013). (2011). stereotactic surgery: preliminary evaluation of 60. Maroof, A. M. et al. Directed differentiation and 80. Lister, R. et al. Hotspots of aberrant epigenomic targeting accuracy. Neurosurgery 70, 95–103 (2012). functional maturation of cortical interneurons from reprogramming in human induced pluripotent stem 102. Draft guidance for industry: considerations for the human embryonic stem cells. Cell Stem Cell 12, cells. Nature 471, 68–73 (2011). design of early-phase clinical trials of cellular and gene 559–572 (2013). 81. Warren, L. et al. Highly efficient reprogramming to therapy products. US Department of Health and 61. Yuan, S. H. et al. Cell-surface marker signatures for the pluripotency and directed differentiation of human Human Services, The Food and Drug Administration isolation of neural stem cells, glia and neurons derived cells with synthetic modified mRNA. Cell Stem Cell 7, [online], http://www.fda.gov/downloads/ from human pluripotent stem cells. PLoS ONE 6, 618–630 (2010). biologicsbloodvaccines/ e17540 (2011). 82. Yu, J. et al. Human induced pluripotent stem cells free guidancecomplianceregulatoryinformation/guidances/ 62. Dubois, N. C. et al. SIRPA is a specific cell-surface of vector and transgene sequences. Science 324, cellularandgenetherapy/ucm359073.pdf (2013). marker for isolating cardiomyocytes derived from 797–801 (2009). human pluripotent stem cells. Nature Biotech. 29, 83. Fusaki, N., Ban, H., Nishiyama, A., Saeki, K. & Acknowledgements 1011–1018 (2011). Hasegawa, M. Efficient induction of transgene-free The authors’ own work described in this Review was sup- 63. Paige, S. L. et al. A temporal chromatin signature in human pluripotent stem cells using a vector based on ported by New York State Stem Cell Board (NYSTEM) human embryonic stem cells identifies regulators of Sendai virus, an RNA virus that does not integrate into (C028503) and the US National Institute of Neurological cardiac development. Cell 151, 221–232 (2012). the host genome. Proc. Jpn Acad. Ser. B Phys. Biol. Disorders (5R01NS054009 to V.T.; NS052671 and 64. Rada-Iglesias, A. et al. Epigenomic annotation of Sci. 85, 348–362 (2009). NS084334 to L.S.). enhancers predicts transcriptional regulators of human 84. Polo, J. M. et al. Cell type of origin influences the neural crest. Cell Stem Cell 11, 633–648 (2012). molecular and functional properties of mouse induced Competing interests statement 65. Arbab, A. S., Liu, W. & Frank, J. A. Cellular magnetic pluripotent stem cells. Nature Biotech. 28, 848–855 The authors declare no competing interests. resonance imaging: current status and future prospects. (2010). Expert Rev. Med. Devices 3, 427–439 (2006). 85. Kim, K. et al. Donor cell type can influence the 66. Budde, M. D. & Frank, J. A. Magnetic tagging of epigenome and differentiation potential of human FURTHER INFORMATION therapeutic cells for MRI. J. Nucl. Med. 50, 171–174 induced pluripotent stem cells. Nature Biotech. 29, California Institute of Regenerative Medicine — Turning (2009). 1117–1119 (2011). stem cells into therapies: http://www.cirm.ca.gov/ 67. Kretzschmar, K. & Watt, F. M. Lineage tracing. Cell 86. Kim, K. et al. Epigenetic memory in induced our-progress/stem-cells-therapies 148, 33–45 (2012). pluripotent stem cells. Nature 467, 285–290 Center for Stem Cell Biology: http://www.mskcc.org/ 68. Kiuru, M., Boyer, J. L., O’Connor, T. P. & Crystal, R. G. (2010). research/stem-cell-biology Genetic control of wayward pluripotent stem cells and 87. Bar-Nur, O., Russ, H. A., Efrat, S. & Benvenisty, N. Guidelines for the clinical translation of stem cells: their progeny after transplantation. Cell Stem Cell 4, Epigenetic memory and preferential lineage-specific http://www.isscr.org/home/publications/ClinTransGuide 289–300 (2009). differentiation in induced pluripotent stem cells New York state stem cell consortia: http://www.mskcc.org/ 69. Björklund, A. et al. Mechanisms of action of derived from human pancreatic islet β cells. Cell Stem new-york-state-stem-cell-science-consortia intracerebral neural implants: studies on nigral and Cell 9, 17–23 (2011). StemBook protocols: http://www.stembook.org/protocols striatal grafts to the lesioned striatum. Trends 88. Cyranoski, D. Stem cells cruise to clinic. Nature 494, ALL LINKS ARE ACTIVE IN THE ONLINE PDF Neurosciences 10, 509–516 (1987). 413 (2013).

92 | FEBRUARY 2014 | VOLUME 15 www.nature.com/reviews/genetics