Regeneration is a regulative developmental process ubiquitous across all kingdoms of life that functions throughout the life cycle to maintain or restore normal form and function of cells, tissues, and in some cases organs, appendages, and whole organisms (Stocum,2006; Carlson,2007; Birnbaum and Sanchez-Alvarado,2008; Edwards,2008, for reviews). The roots, stems and leaves of plants have extensive regenerative capacity, and entire plants can grow from single cells or small cuttings. The greatest powers of animal regeneration are found in invertebrates, such as hydra, planaria, and ascidians, which can remodel small fragments of the body into miniature wholes capable of feeding and growing to the original size, or regenerate the missing parts of the body from an amputated part. Among the vertebrates, some amphibians can regenerate amputated tails, jaws, and appendages (Stocum,2006, for review), and even the distal tips of mammalian digits, including humans, are capable of regeneration (Illingworth,1974; Borgens,1982; Han et al.,2003,2008). The patterning of a whole body from a fragment, or of a part regenerating from a remainder, is accomplished by the reorganization or addition of positional identities by intercalation within a boundary system (Agata et al.,2007).
The regenerative capability of most vertebrate animals is restricted to certain tissues. In the absence of injury, many cell types, such as epithelia and blood cells, turn over rapidly, while others, such as hepatocytes, myofibers, osteocytes, and most neurons, have low turnover rates, or do not turn over at all. In organisms that grow throughout life, such as fish, the total number of cells in various tissues increases continuously (Zupanc and Horschke,1995), indicating that the number of new cells produced is higher than the number of cells lost. The replacement of cells in the absence of injury is called homeostatic regeneration because it precisely maintains the cell number appropriate for the mass and architecture of a tissue. By contrast, the loss of normal tissue mass and/or architecture to acute injury or disease requires a quantitatively more intense and sometimes qualitatively different regenerative response that restores the tissue to its original state. This response is called injury-induced regeneration.
All organs of the body contain epithelial, endothelial, and connective tissue components. Although the epithelial and endothelial components regenerate readily, injuries to the connective tissue component above a critical threshold can induce fibrosis (scarring). Whether a connective tissue component of an injured organ produces scar or regenerates is a function of cellular interactions that create either a regeneration-permissive or inhibitory “niche” environment consisting of specific sets and relative concentrations of extracellular matrix (ECM) and cell surface-associated and soluble ligands, such as growth factors and cytokines (Naverias and Daley,2006, for review). For example, injury to the dermis of the skin or the interstitial tissue of the myocardium results in inflammation and secretion by fibroblasts of collagen fibers that are remodeled into cross-linked bundles characteristic of scar tissue (Linares,1996, for review). By contrast, injury to the connective tissue of liver does not result in fibrosis, and it regenerates (Michalopoulos and DeFrances,1997, for review).
Animals use three basic mechanisms of regeneration (Fig. 1). The first is compensatory hyperplasia, the proliferation of cells in their differentiated state. The second and most common mechanism is the activation and proliferation of adult stem cells (ASCs), undifferentiated cells residing in regenerating tissues. The third is dedifferentiation, a process by which differentiated cells lose their phenotypic specializations, adopt a developmentally earlier pattern of gene activity, and proliferate. Table 1 summarizes the organisms and tissues that regenerate by each of these three mechanisms. No examples of compensatory hyperplasia exist in plants, but plants regenerate by means of ASCs in the form of meristems and pericycle founder cells, and by the dedifferentiation of mature cells to callus (Birnbaum and Sanchez-Alvarado,2008, for review).
Table 1. Mechanisms of Regeneration Used by Various Animals
Although liver and pancreas are normally regenerated via compensatory hyperplasia, they also harbor stem cells that are activated when the proliferative ability of their differentiated hepatocytes and islet cells is compromised.
Liver,a pancreas,a blood, muscle, bone, epithelia, some neurons
Vertebrates (fish, amphibians)
Lens, retina, spinal cord, jaws, limbs and fins, tails
The adult stem cells of mesodermally derived tissues in animals are mesenchymal in nature, meaning that they have the ability to migrate on substrates. By contrast, epithelial adult stem cells adhere to one another at their basolateral surfaces by tight junctions, adherens junctions, and gap junctions, and to a basement membrane by hemidesmosomes. To regenerate damaged epithelium, epithelial stem cells must migrate to bridge gaps. To migrate, the cells undergo epithelial to mesenchymal transformation (EMT), in which they eliminate their specialized attachments and take on mesenchymal characteristics. Once the gap has been bridged, the mesenchymal cells undergo the reverse mesenchymal to epithelial transformation (MET) to restore the epithelium (Chernoff et al.,2003; Zeisberg and Kalluri,2004).
Over the past 5 years, regenerative biology and medicine has emerged as one of the potentially most revolutionary fields of 21st century life sciences (Stocum,2006; Carlson,2007; Edwards,2008; Atala et al.,2008, for reviews). In just the past year alone, a torrent of basic and clinical research papers has flooded the literature with exciting discoveries and insights. The primary focus of regenerative medicine is rebuilding damaged tissues by transplanting cell suspensions or aggregates into the locus of injury (Daley and Scadden,2008, for review). This approach has been used in blood transfusions since the first decade of the 20th century, and since 1968 in the technically more demanding procedure of bone marrow transplantation to reconstitute the hematopoietic system of persons who have undergone chemotherapy or total body irradiation for malignancies, or who suffer from genetic diseases of the hematopoietic system (Garovoy et al.,1997, for review). A variation of this approach is to make an implantable bioartificial tissue by seeding cells into a decellularized natural ECM, or a polymer, metal, or ceramic scaffold (Nerem and Sambanis,1995; Hubbell,2004).
A major issue for cell transplant therapies is the source of the cells to be used. Three sources of cells can be tapped for transplant (Fig. 2): differentiated tissues, ASCs, and derivatives of embryonic stem cells (ESCs). Each of these could be autogeneic (or syngeneic), allogeneic, or xenogeneic. Furthermore, they could be wild-type, mutant, or chemically modified in some way, as well as being freshly prepared or cultured for some period of time. The ideal cell source would be easy to access, have high growth potential, be pluripotent, and not be subject to immunorejection. Differentiated cells can be autogeneic but do not meet the pluripotency criterion, and most are difficult to expand in vitro, although differentiated articular cartilage cells expanded in culture are in clinical use to repair partial thickness defects in articular cartilage of the knee joint (Brittberg et al.,1999). The primary focus of cell transplant therapies has therefore been ASCs and ESCs.
In this review, we explore four topics of regenerative science and medicine. The first compares the current level of performance of ASCs and ESCs in animal and human cell transplant therapies, and the drawbacks of each for transplant therapy. Because developmental plasticity and immunorejection are major issues in the use of allogeneic or xenogeneic ASCs or ESC derivatives, the second topic explores somatic cell nuclear transfer (SCNT) as a method to derive autogeneic, pluripotent ESCs. The third topic explores how SCNT and ESC research has led to the ability to derive pluripotent, autogeneic ESCs by reprogramming adult somatic cells. Lastly, we discuss in a Perspective how research on somatic cell reprogramming can evolve regenerative medicine from a platform consisting solely of cell transplantation to one that includes the chemical induction of regeneration from the body's own cells at the site of injury.
ADULT STEM CELLS VERSUS EMBRYONIC STEM CELLS AS REGENERATIVE THERAPIES
Adult Stem Cells
ASCs regenerate epithelia, brain tissue, muscle, blood, and bone. They have also been found in other tissues that normally scar after injury, such as myocardium, spinal cord, and retina (Stocum,2008, for review). Here we focus on the ASCs of central nervous system (CNS), cardiac, and musculoskeletal tissues.
Regeneration of central nervous tissue.
In the mammalian brain, neural stem cells (NSCs) are found in two areas. First is the anterior part of the subventricular zone of the lateral ventricle, from where the immature neurons migrate by means of the so-called rostral migratory stream into the olfactory bulb (Altman,1969; Luskin,1993; Lois and Alvarez-Buylla,1994; Pencea et al.,2001). These new cells differentiate predominantly into granule neurons and—to a lesser extent—periglomerular interneurons. Second is the subgranular zone of the dentate gyrus, from where the new cells migrate a short distance into the granule cell layer of the hippocampus (Altman and Das,1965; Kaplan and Bell,1984; Eriksson et al.,1998; Gould et al.,1999; Kornack and Rakic,1999; Seri et al.,2001). These cells develop into mature granule neurons. The neurogenic potential of such adult stem cells could form the basis for a replacement therapy for neurons lost to injury or neurodegenerative disease (Zhao et al.,2008, for review). However, although brain injuries lead to increased proliferation of progenitors in the subgranular and subventricular zones, and even appear to stimulate proliferation of quiescent neural progenitors in regions where adult neurogenesis is absent in the intact brain (Magavi et al.,2000; Chen et al.,2004), the total number of new neurons is a tiny fraction of the number required to replace the degenerated neurons, and most of the new cells fail to survive and/or differentiate. For example, after experimentally induced stroke in rodents, immature neurons born in the subventricular zone migrate to the damaged striatal area. Although these cells start to express markers for striatal medium-sized spiny neurons (the phenotype most severely affected by the insult), and some survive for over 1 year (Chen et al.,2004), the majority of them die within a few weeks (Arvidsson et al.,2002).
Intense effort has been put into developing NSC-based therapies for spinal cord injury and neurodegenerative disease (Singec et al.,2007). Allogeneic NSCs transplanted to spinal cord lesions have been reported to promote partial recovery from paralysis. However, the modest improvements reported are likely not due to the differentiation of donor NSCs into new neurons, but to effects on host cells. NSCs transplanted into the spinal cord of rodents differentiate primarily into glial cells, including oligodendrocytes (Hofstetter et al.,2005; Karimi-Abdolrezaee et al.,2006). Rats injected intravenously with human fluorescein isothiocyanate-labeled umbilical cord blood 5 days after lesioning the spinal cord were reported to achieve partial recovery of locomotory behavior (Saporta et al.,2003). Histological examination indicated that the labeled cells, of which fewer than a thousand survived, did not differentiate into either neurons or glia. Infusions of bone marrow cells into a group of 32 patients at the University of Sao Paulo 2–12 years after complete spinal cord injury were reported to modestly improve lower extremity function in 15 patients (see Steeves et al.,2004). These results suggest that the modest improvements observed are likely due to axon remyelination by grafted cells, and/or paracrine/juxtacrine effects of the transplanted cells on host neurons.
Direct evidence for paracrine/juxtacrine effects of transplanted cells on host neural tissue has been obtained. Mikami et al. (2004) reported that splenic dendritic cells transplanted into lesioned mouse spinal cords activated the proliferation and differentiation of host NSCs into new neurons and induced axon sprouting, accompanied by partial recovery from hindlimb paralysis. Co-culture of spinal cord NSCs with dendritic cells significantly enhanced the survival and proliferation of the NSCs. By contrast, medium conditioned by dendritic cells had only one-tenth the enhancing activity observed in co-cultures, indicating that the major effect of the dendritic cells on neurons is mediated by cell contact. Dendritic cells secreted NT-3 in vitro and in vivo, so the minor enhancing activity of conditioned medium may have been exerted through this trophic molecule. Human mesenchymal stem cells (MSCs) implanted into the dentate gyrus of the mouse hippocampus promoted neurogenesis by endogenous NSCs (Munoz et al.,2005) and astrocytes derived from embryonic glial-restricted precursors transplanted into transection lesions of the rat spinal cord promoted axon regrowth and suppressed initial scarring that was associated with significant improvement of locomotor function (Davies et al.,2006). Kang et al. (2006) reported that oligodendrocyte precursors differentiated from adipose-derived stromal cells promoted functional recovery by both remyelination and stimulation of proliferation and differentiation of host NSCs when grafted into rat spinal cord lesions.
Regeneration of myocardium.
Recent reviews have summarized the effects of clinical trials in which satellite cells (SC, the stem cell of skeletal muscle) and bone marrow cells (BMC) were transplanted to ameliorate the effects of myocardial infarction (Togel and Westenfelder,2007; Segers and Lee,2008; Passier et al.,2008, for reviews). The results have been generally disappointing, with improvements in cardiac function ranging from nonexistent to modest. Neither SCs nor BMCs differentiated into new cardiomyocytes. The SCs differentiated into skeletal muscle that was not electrically coupled to host cardiac muscle, but may have augmented contraction of the heart muscle. Bone marrow cells might have had a positive effect in two ways, through endothelial progenitors contributing to new microvasculature, and by paracrine/juxtacrine effects on survival of host cardiomyocytes. Paracrine effects are supported by the results of experiments in which MSCs transfected with the Akt gene (which enhances the survival of MSCs under hypoxic conditions) were injected into myocardial infarctions of mice (Mangi et al.,2003). The area of infarction was reduced to zero and left ventricular ejection fraction was restored to normal. Gnecchi et al. (2005) then showed that conditioned medium of MSC cultures transfected with Akt increased the survival of cardiomyocytes under hypoxic conditions in vitro and in vivo. One paracrine agent responsible for these effects is likely to be thymosin β4, which has been shown in mice to be up-regulated after myocardial infarction. Thymosin β4 protects cardiomyocytes from hypoxia and stimulates angiogenesis from stem cells derived by EMT of epicardium (Bock-Marquette et al.,2004; van Tuyn et al.,2007; Smart et al.,2007; Mani et al.,2008). The intact thymosin β4 molecule induces significant outgrowth of stem cells from epicardial explants and a cleavage product of thymosin β4, N-acetyl-seryl-aspartyl-lysyl-proline, promotes their differentiation into fibroblasts, smooth muscle cells and endothelial cells. An ECM molecule, periostin, is also up-regulated after myocardial infarction, and when delivered exogenously has been shown to aid the repair of infarcted rat myocardium by promoting cardiomyocyte division (Kuhn et al.,2007).
The myocardium contains several populations of cardiac stem cells that differ in c-kit/Sca-1 expression (Anversa et al.,2007; Segers and Lee,2007, for reviews). One population of these cells has been reported to regenerate the myocardial wall in the rat heart (Beltrami et al.,2003), but others have not been tested for their in vivo regenerative capacity. There is some skepticism as to whether these cells are actually endogenous stem cells or bone marrow cells patrolling for toxic molecules in peripheral tissues (Passier et al.,2008, for review). Hsieh et al. (2007) tested the idea of cardiomyocyte renewal from stem cells by marking differentiated mouse cardiomyocytes with an inducible transgene for green fluorescent protein (GFP). After myocardial infarction and induction of the GFP reporter with tamoxifen, approximately 15% of the GFP-labeled cardiomyocytes was replaced by unlabeled cardiomyocytes. These results suggest that new cardiomyocytes can indeed be produced from cardiac stem cells.
Regeneration of musculoskeletal tissues.
There is a large body of literature on the use of satellite cells and mesenchymal stem cells to treat muscular dystrophy, and cartilage and bone defects (Stocum,2006, for review). Allogeneic SC transplants to regenerate normal muscle in place of dystrophic muscle in mdx mice and muscular dystrophy patients have not been successful due to inflammation, immune rejection, poor survival and limited migration of the transplanted cells. Another problem is that expansion of SCs in vitro is associated with a reduction in their regenerative capacity due to loss of proliferation potential and a tendency to differentiate more rapidly (Montarras et al.,2005). Recently, a subpopulation of SCs has been identified that, when freshly purified and injected into mdx mouse muscles, significantly restores dystrophin expression and muscle function (Cerletti et al.,2008). These cells, termed skeletal muscle precursors, also became self-renewing satellite cells able to participate in sequential rounds of regeneration. In another promising approach, Benchaouir et al. (2007) introduced dystrophic human myogenic progenitor cells (CD133+) genetically corrected by forced exon skipping (Goyenvalle et al.,2004) into the muscles of scid/mdx mice. The mice showed a significant recovery of dystrophin-positive myofibers and muscle function.
Because MSCs normally give rise to chondroblasts and osteoblasts, they are logical candidates to repair defects in articular cartilage and bone. Defects in articular cartilage that penetrate to the subchondral bone heal with fibrocartilage due to the seepage of blood and MSCs into the defect, whereas there is no regeneration at all in defects that involve the hyaline cartilage alone (Caplan et al.,1993). The results of transplanting MSCs alone or in scaffolds into defects that involve only the hyaline cartilage have been generally disappointing. The transplanted MSCs may initially form hyaline cartilage, but the cartilage does not integrate with surrounding host cartilage, develops fissures, and thins over time (Gelse et al.,2003; Tibesku et al.,2004; Djouad et al.,2006, Vilquin and Rosset,2006; Stocum,2006, for reviews).
Infants with osteogenesis imperfecta, or “brittle bone” disease, in which a mutation in the type I collagen gene makes the bone matrix highly susceptible to fracture, have been successfully treated by the engraftment of normal bone marrow cells (Horwitz et al.,1999), and grafts of autogeneic bone marrow cells have promoted the healing of nonunion fractures (Connolly,1995). The MSCs of the bone marrow are likely responsible for rebuilding new bone tissue. Bioartificial bone segments, made by seeding culture-expanded MSCs into a variety of osteoconductive and inductive scaffolds, with and without added growth factors, have been partially successful in replacing long bone segments in small and large animal models (Caplan et al.,1993; Cancedda et al.,2007, for reviews). Cultured bone marrow cells and adipose-derived adult stromal cells seeded into an apatite-coated poly (lactic-co-glycolic acid; PLGA) scaffold completely repaired 5-mm circular defects in mouse parietal bone in the absence of any exogenous growth factors (Cowan et al.,2004). Labeling experiments showed that nearly 100% of the osteoblasts in the new parietal bone were derived from donor cells. In a spectacular clinical application, a bioartificial mandible was fabricated by putting together a titanium scaffold seeded with bovine bone matrix, autogeneic bone marrow cells, and BMP-2, and culturing the construct in a space made in the latissimus dorsi muscle of the patient. The mandible was then implanted in place of the original, which had been removed 9 years earlier because of oral cancer (Warnke et al.,2004). The mandible functioned properly during most of the interval until the patient's death from cardiac arrest 15 months later, but revealed several problems to be addressed, including the nondegradability of the titanium scaffold and a need for enhancing bone induction for more homogeneous mineralization (Warnke et al.,2006).
To summarize, what we have learned from the studies cited and many others, is that adult stem cell therapy has real potential to regenerate at least muscle and bone damaged by injury or genetic disease, and that cardiac stem cells may be a way to regenerate new cardiomyocytes after myocardial infarction. A significant lesson learned from experiments and clinical trials to treat myocardial infarction and CNS injury is that a major effect of transplanted ASCs (and other cells) is the production of paracrine factors and juxtracrine contacts that have protective and regeneration-stimulatory effects on host tissues. Such effects argue for the development of a chemical induction approach to regenerative medicine, discussed in the Perspective section of this article.
Embryonic Stem Cells
Embryonic stem cells are cultures of the inner cell mass (ICM) cells of the blastocyst or epiblast of mammalian embryos, or equivalent stages of other embryos (Smith,2001; Rippon and Bishop,2004, for reviews). ESCs express species- and stage-specific embryonic surface antigens (SSEAs), alkaline phosphatase, and high levels of telomerase. The intense interest in ESCs for regenerative medicine lies in the fact that they have been proven to be pluripotent (in the sense that they can differentiate into tissues of all three embryonic germ layers) and can be expanded indefinitely in vitro. Human ESC cultures were established a decade ago from unused in vitro fertilization blastocysts retained in assisted reproduction facilities (Thomson et al.,1998), and from primordial germ cells of 5–9 week spontaneously aborted embryos (Shamblott et al.,1998). ESCs are made by dissociating ICM cells and growing them on a layer of irradiated feeder fibroblasts in the presence of leukemia inhibitory factor (mouse) or fibroblast growth factor (FGF)-2 (human) to maintain the cells in a pluripotent, self-renewing state. Protocols for differentiating ESCs to ASC-like or terminal cell phenotypes, based on what is known about the developmental programs of each cell type, have been devised for a wide variety of cell types (Murry and Keller,2008, for review). In addition, screens of combinatorial chemical libraries have identified several synthetic small molecules that induce the differentiation of neural cells and cardiomyocytes from hESCs (Xu et al.,2008, for review).
Regulatory circuitry for pluripotency.
The regulatory circuitry for pluripotency is established during fertilization and preimplantation development of the zygote (Fig. 3). Before fertilization, the haploid sperm chromatin is highly compacted by protamines, which upon fertilization are exchanged for histones from the maternal cytoplasmic pool, relaxing the chromatin (Morgan et al.,2005, for review). The maternal genome is already packaged with histones before fertilization, and is arrested in metaphase of the second meiotic division. At fertilization the egg completes the second meiotic division to achieve haploidy. As the zygote develops through morula to blastocyst stages, genes are activated that establish and maintain the pluripotency of ICM cells and also repress differentiation genes (Morgan et al.,2005; Jaenisch and Young,2008; Silva and Smith,2008; Yamagata,2008, for reviews).
The central unit of chromatin is the nucleosome, an octet of eight histone proteins wrapped with short segments of DNA. The histone proteins interact with the DNA to regulate its transcriptional activity. During development, the nucleosomes and the DNA are modified epigenetically in ways that specify the gene activity patterns for cell differentiation (Morgan et al.,2005, for review). The nucleosomes are physically rearranged on the DNA by complexes of chromatin-remodeling proteins. Their histones are methylated/demethylated or acetylated/deacetylated on specific amino acids, and the DNA itself is methylated/demethylated at CpG (cytidine-phosphate-guanine) islands by histone and DNA methyltransferases/demethylases and histone acetyltransferases and deacetylases. Combinations of these epigenetic “marks” and nucleosome positions determine the degree of compaction of the chromatin and thus its transcriptional activity (Morgan et al.,2005; Bibikova et al.,2008, for reviews). The epigenetic marks are heritable during cell division.
An interaction network of over 35 proteins that initiates and maintains pluripotency has been mapped for ES cells (Wang et al.,2006). A trio of transcription factors, Oct4, Sox2, and Nanog lies at the heart of this network (Silva and Smith,2008; Jaenisch and Young,2008, for reviews). Oct4 and Sox2 are the critical core factors for pluripotency, but they also direct the production of FGF4, which promotes the spontaneous differentiation of ESCs by Erk signaling. Nanog is not essential to initiate pluripotency, but appears to stabilize the pluripotent state induced by Oct4 and Sox2 by countering the effects of Erk signaling (Silva and Smith,2008). The “three amigos” and their associated proteins of the pluripotency network regulate three sets of genes (Fig. 3). First, they bind to their own and each other's promoters to autoregulate their own production. Second, they bind to the promoters of genes for transcription factors (STAT3, HESX1), cell signals (Lefty2, FGF2), and chromatin-modifying enzymes that promote ESC self-renewal. Third, they bind to the promoters of a large group of genes that direct the differentiation of ectodermal, mesodermal, and endodermal derivatives, allowing them to be silenced by Polycomb group (PcG) proteins that catalyze the methylation of lysine 27 on histone 3 (H3K27me; Orlando,2003, for review). Another nuclear protein unrelated to known master regulators, but essential for maintenance of pluripotency, is Ronin (masterless Japanese samurai). Ronin binds directly to host cell factor 1 protein, a key regulator of transcriptional control associated with protein complexes that epigenetically modify histones, and its overexpression partially negates the requirement for Oct4 to maintain pluripotency (Dejosez et al,2008).
Specific subsets of signaling factors promote ESC differentiation into a wide variety of cell types (Silva and Smith,2008, for review). Commitment to differentiation involves silencing Oct4 and activating differentiation genes by relieving Polycomb-mediated gene repression through the action of UTX and JMJD3 demethylases (members of the jumonji domain proteins) and the methylation of H3K4 (Trojer and Reinberg,2006; Swigut and Wysocka,2007, for reviews; Fig. 3). Retinoic acid, an inducer of neural differentiation, directly contributes to Oct4 silencing by activating nuclear receptors that bind to retinoic acid receptor elements in the Oct4 promoter, and nuclear repressors are induced that are required for silencing. Histone modifications associated with active transcription (H3K4me3, H3K7, and H3K9 acetylation) are lost at Oct4, while modifications associated with heterochromatin and repressed transcription (H3K9me2/3) are gained (Jaenisch and Young,2008, for review). In addition, the promoters of Oct4 and other ESC-specific genes, including Rex1 (but not Nanog or Sox2), undergo DNA methylation. Micro RNAs (miRNAs) are also part of the regulatory network of pluripotency (Ivey et al.,2008, for review). A subset of miRNAs is preferentially expressed in ESCs. ES cells deficient in miRNA processing enzymes do not proliferate normally, fail to completely down-regulate Oct4 expression, and exhibit defects in differentiation (Stadler and Ruohola-Baker,2008).
Regeneration of central nervous tissue.
The first clinical treatment of a neurodegenerative disease by cell transplantation aimed at a functional recovery from Parkinson's disease through restoration of dopaminergic neurotransmission in the nigrostriatal pathway (for review, see Lindvall,1989). This disease affects an estimated one million persons in the United States alone (see www.pdf.org). Mesencephalic dopaminergic neurons from human embryos are grafted to replace dopamine-expressing nigrostriatal neurons lost to disease. However, despite a large number of clinical trials since the 1980s, the effectiveness of the transplantation approach for Parkinson's has been difficult to assess, due to the low number of randomized trials, the lack of proper controls, the too short follow-up examinations, and the inappropriate choice of outcome measures (Stowe et al.,2003; Nguyen,2004). In the few studies in which a sufficient number of patients were used, and the patients were randomly assigned to receive a transplant or undergo sham surgery, limited improvement was observed in younger patients, but no improvement was seen in patients older than 60 years (Freed et al.,2001).
ESCs have now been tested as regenerative therapies for a variety of tissues, including Parkinson's disease. After transplantation into the striatum of a rodent model in which loss of midbrain dopamine neurons was induced by administration of 6-hydroxydopamine, both undifferentiated ESCs (Björklund et al.,2002), as well as ESCs that had been directed to differentiate into dopaminergic neurons (Kim et al.,2002; Barberi et al.,2003), survived for at least 16 weeks and displayed morphological, molecular, and functional features consistent with those of normal dopamine-expressing nigrostriatal neurons. Furthermore, animals with grafts showed gradual and sustained improvement of motor dysfunction. However, of 25 rats that had received undifferentiated ESCs, 6 rats showed no graft survival, and 5 other rats developed teratoma-like tumors at the implantation site (Björklund et al.,2002).
McDonald et al. (1999) transplanted ESC-derived mouse neural/glial precursors into 9-day-old rat spinal cord lesions. Staining with antibodies specific for mouse proteins and for glial and neuronal markers showed that many of the implanted cells survived, migrated throughout the injured area and differentiated into new interneurons, oligodendrocytes, and astrocytes. These findings were correlated with regaining the ability to bear weight on their hind legs and by restoration of partly coordinated stepping movements, suggesting that signal transmission was partially restored between brain and hind legs.
Injection of oligodendrocyte precursors derived by the directed differentiation of mouse ESCs have also proven to be effective in CNS remyelination in a rat model of the demyelination disorder Pelizaeus-Merzbacher disease, which is caused by a mutation in the X-linked gene for myelin proteolipid protein (Brüstle et al.,1999) and after injection of oligodendrocyte precursors derived from human ESCs in the shiverer mouse, which is homozygous for a mutation in the myelin basic protein gene (Nistor et al.,2005; Windrem et al.,2008).
Regeneration of myocardium.
ESC-derived cardiomyocytes have also been tested as therapy for myocardial infarction in animals (Van Laake et al.,2005, for review). Mouse ESC-derived cardiomyocytes were stably integrated into the ventricular myocardium of mdx dystrophic mice after transplantation (Klug et al.,1995,1996). When injected into the left ventricle of pigs in which a complete atrioventricular block had been induced by ablating the bundle of His, human ESC-derived cardiomyocytes restored normal electrical rhythm (Kehat et al.,2004). Immunostaining with anti-human mitochondrial antibodies confirmed the presence of human cardiomyocytes in the hearts that were integrated with host cells. Human ESC-derived cardiomyocytes engraft and improve cardiac function in the infarcted hearts of rats (Laflamme et al.,2007; Van Laake et al.,2005). Cardiac function decreased after 12 weeks in the study of van Laake et al. (2005). Laflamme et al. (2007) used a cardioprotective molecular cocktail to enhance survival of the grafted cardiomyocytes, but their study was terminated after 4 weeks; thus it remains unclear whether cardioprotection was maintained (see Rubart and Field,2007). Furthermore, several investigators have reported that transplanted ESC-derived cardiomyocytes never differentiate to a mature phenotype and resemble fetal cardiomyocytes. Thus far, no clinical trials transplanting ESC-derived cardiomyocytes have been attempted.
Regeneration of β-cells and dystrophic muscle.
Progress is also being made toward the use of ESCs to derive functional β-cells for treatment of diabetes (Otonkoski et al.,2005; Kume,2005; Spence and Wells,2007, for reviews) and myogenic cells for treatment of muscular dystrophy. D'Amour et al. (2005) developed a five-step protocol to direct the differentiation of human ESCs to pancreatic islet cells, including insulin-producing cells. These cells differentiate into over 50% β-cells when implanted into mice, produce human insulin and C-peptide in response to glucose stimulation, and protect against streptozotocin-induced hyperglycemia (Kroon et al.,2008). Protocols have also been developed to differentiate and purify a myogenic cell population from mouse ESCs that has significant potential for muscle regeneration. These cells were able to engraft into mdx mouse muscle and form myofibers (Darabi et al.,2008).
In summary, ESCs show great promise as a cell source for the regeneration of new tissue, due to their high growth and self-renewal capacity, and their ability to differentiate into a myriad of precursor or differentiated cell types when directed by the appropriate set of environmental factors.
Issues Surrounding the Use of ASCs and ESCs for Cell-Based Therapy
Several problems currently limit the potential of ASCs for regenerative therapies. First, it is generally difficult to get ASCs in the numbers required for cell therapy, because they are few in number in vivo, difficult to harvest, and tend to differentiate in vitro, making them difficult to expand. A better understanding of the composition and three-dimensional spatial organization of ASC microniches (“stem cell ecology”; Powell,2005) will be essential to achieving their expansion in vitro without loss of regenerative potential. Where the injury environment is not adequate to support regeneration in the presence of regeneration-competent cells, for example in the spinal cord or myocardium, it will be essential to determine what molecular additives will establish a microniche that supports regeneration and inhibits scarring. Another way to generate large numbers of ASCs is to immortalize them by inhibiting checkpoints in the cell cycle. For example, Roy et al. (2004) reported the immortalization of human neural progenitor cells by use of a retrovirus encoding hTER, the rate-limiting component of the telomerase enzyme complex. The cells matured as neurons when grafted to the fetal rat brain and injured adult spinal cord, showing that immortalization does not affect their ability to redifferentiate, although it is unknown how long the neurons survived.
Second, we would much prefer to have a pluripotent cell source that would be a “universal donor” with regard to its developmental potency to meet patient need. There has been a great deal of interest in whether or not adult stem cells might have developmental potency that exceeds their normal developmental fate when placed in niches they normally would never see. Lineage conversion has been quite clearly observed during tail regeneration in salamander larvae (Escheverri and Tanaka,2002). GFP-labeled ependymal cells of the spinal cord transformed into cartilage, muscle, and other mesodermal cell types at a frequency of better than 10%. However, the results of assays to test the plasticity of mammalian bone marrow and other ASCs have been inconsistent and controversial. The most widely used assay for pluripotency has been the lethally irradiated mouse. This assay has suggested in some experiments that hematopoietic and mesenchymal stem cells can change their prospective fates at low frequency, but other experiments suggest that these changes are illusions due to fusion with host cells (which itself could be a repair mechanism), problems with labeling, or cellular mimicry, a phenomenon in which labeled cells are incorporated into a tissue, display some molecular markers of the differentiated cells of that tissue, but do not survive (Rizzino,2007, for review). In general, there seems to be an over-reliance on molecular markers as indicators of differentiation. Such markers can be misleading if they are not backed up by evidence of morphological and functional differentiation (Jaenisch and Young,2008). The most stringent assay for pluripotency is the ability of cells to participate in the formation of late stage or live birth chimeric embryos after injection into a blastocyst. However, this assay has been used only twice, on NSCs, with opposite results (Clarke et al.,2000; D'Amour and Gage,2003).
ESCs, on the other hand, are clearly pluripotent and can be expanded indefinitely in an undifferentiated and self-renewing state. Nevertheless, ESCs present several problems for cell transplant therapy (Deb and Sarda,2008; Daley and Scadden,2008, for reviews). First is the possibility of contamination of the ESCs from mouse feeder layers or animal serum-supplemented medium with pathogens or xenogens (cells or proteins) that might trigger host immune reactions. Protocols have been developed to grow hESCs on animal-derived ECM (Xu et al.,2001) or on human ECM in medium conditioned by mouse feeder cells (Amit et al.,2004). Newer protocols are being developed to grow hESCs in chemically defined medium without animal sera (Carpenter et al.,2004). In one such protocol, hESCs are grown on ECM of mouse embryonic fibroblasts in a defined medium supplemented with a serum replacement containing animal proteins (Klimanskaya et al.,2005). Replacing these with human proteins would be desirable. The optimum defined medium and substrate for growing hESCs will require identification of the factors produced by human feeder cells that help maintain the cells in an undifferentiated state and promote their proliferation.
Second, undifferentiated human ESCs form benign teratomas when injected subcutaneously or intramuscularly into SCID mice (Thomson et al.,1998). Directed differentiation protocols currently can produce differentiated cell populations that are no more than 80% pure, whereas for clinical use the purity must be 100% to avoid the possibility of teratoma formation. Complete purification might be achieved by fluorescence activated cell sorting (FACS), or treating the cells with N-oleoyl serinol, which selectively induces apoptosis of any undifferentiated ESCs (Bieberich et al.,2004). Furthermore, it is currently possible to direct the differentiation of only relatively small quantities of cells from ESCs. The future use of ESC derivatives for tissue regeneration will depend largely on the development of differentiation protocols that can produce large quantities of robustly and stably differentiated cells or precursors.
Third, the epigenetic modifications of early development are vulnerable to environmental or culture conditions in in vitro fertilization embryos, so the ESCs derived from them may be epigenetically unstable. A similar epigenetic instability can occur over time in culture (Allegrucci et al.,2007). Chromosome abnormalities also may accumulate in cultures of ESCs, although the development of such abnormalities seems to depend on whether passaging is done by enzymatic or mechanical methods (Buzzard et al.,2004; Mitalipova et al.,2005).
ASCs and ESCs face some common hurdles as therapies. Two of the most important ones are how to accurately deliver and/or home the cells or their derivatives to the injury site, and immunorejection. Delivery and homing is particularly problematic for the replacement of cells that lie deep in the brain, such as dopaminergic neurons (Svendsen,2008). Either allogeneic or xenogeneic ASCs or ESCs and their derivatives will elicit an immunorejection response when transplanted, necessitating the use of immunosuppressants to avoid rejection. This problem might be solved either by understanding how to induce peripheral tolerance (Waldmann and Cobbold,2004, for review), or by the establishment of banks of ESC lines large enough to provide a wide range of human leukocyte antigen profiles that increase the probability of a close match over a high percentage of a population (Carpenter et al.,2004; O'Rourke et al.,2008).
AVOIDING IMMUNOREJECTION: PRODUCTION OF AUTOGENEIC ESCS BY SCNT
The most direct solution to the immunorejection problem is the derivation of patient-specific (autogeneic) ESCs. This can be accomplished by somatic cell nuclear transfer (SCNT; Fig. 4), in which a patient's somatic cell nucleus is introduced into an enucleated egg, stimulating the egg to undergo mitosis. At the blastocyst stage, the inner cell mass is removed and used to create an autogeneic ESC line. Cells derived from this line will not be immunorejected when transplanted into the nuclear donor because they express the multiple histocompatability genes of the donor. Such autogeneic ESC lines were first created by SCNT for mice and were shown to be capable of differentiating into neurons and muscle in vitro (Munsie et al.,2000; Wakayama et al.,2001). Reproductive cloning using these ESCs as nuclear donors gave rise to fertile adult mice, demonstrating the pluripotency of their nuclei (Wakayama et al.,2001). Recently, autogeneic ESCs that could differentiate into cell types representative of all three germ layers were derived from SCNT blastocysts of monkeys (Byrne et al.,2007), and human SCNT blastocysts were cloned (French et al.,2008). Together, these experiments constitute proof of principle that human autogeneic ESCs can be generated by SCNT.
The chromatin of the donor nucleus for SCNT comes with a full set of differentiated epigenetic marks and nucleosome organization that defines its transcriptional activity. Dedifferentiation of the nucleus requires the erasure of these marks by chromatin modifying enzymes in the egg cytoplasm and acquisition of the marks and nucleosome positions characteristic of zygote chromatin (Fig. 4). As the zygote proceeds through preimplantation development, the chromatin will acquire the epigenetic marks and structure conferring pluripotency through the reactivation of Oct4, Sox2, and Nanog, and by the silencing of differentiation genes. At gastrulation, the marks will recapitulate the epigenetic process of de-repressing differentiation genes, leading to the formation of ectoderm, mesoderm, and endoderm and beyond.
Two very big hurdles to the routine creation of patient-specific ESCs by SCNT are (1) a shortage of human eggs and (2) a low efficiency of reprogramming. While it is possible to do SCNT on fertilized eggs, which are easier to harvest, it is unlikely that this source would alleviate the shortage. Thus, the possibility of using animal eggs as recipients of human somatic nuclei is being explored. Chen has derived ESCs from blastocysts created by fusing enucleated rabbit eggs with somatic cell nuclei from human foreskin and facial skin (see Dennis,2003). The growth potential of these hybrid cells is not clear, and it is possible that they would become unstable due to species incompatibilities between the proteins produced by the human nucleus and those produced by the rabbit maternal mitochondria. In addition to these logistical and biological problems, there are deep bioethical issues about the morality of first making, and then taking, a potential human life for selfish reasons (report of the President's Council on Bioethics,2002; Gilbert et al.,2005; Stocum,2006, for reviews).
An alternative to SCNT for making patient-specific cells would be to fuse a somatic cell to an ESC cytoplast. However, Do and Scholer (2004) reported that hybrids made by fusing mouse ESC karyoplasts to NSCs induced the expression of Oct4 in ESC:NSC hybrid cells, but ESC cytoplasts could not do so. This result suggests that the ESC produces the proteins necessary to confer pluripotency but that these proteins are not stored in the cytoplasm as they are in the mature egg. Cowan et al. (2005) confirmed this by showing that human fibroblasts were reprogrammed to pluripotency after fusion with nucleated human ESCs. However, a major problem with this technique is the tetraploidy of the hybrid cells, which necessitates the technically difficult removal of the ESC nucleus.
REPROGRAMMING OF SOMATIC CELLS TO EMBRYONIC STEM CELLS
Many of the problems of making patient-specific ESCs by SCNT or ESC/somatic cell fusion would disappear if ESCs could be created by chemically inducing the dedifferentiation of somatic cells in vitro. The know-how required to do this, however, was thought to be dauntingly complex, and the ability to do such reprogramming thus years away. However, a surprisingly simple method to reprogram somatic cells has emerged, and has opened a new era of regenerative biology and medicine.
Generation of iPS Cells In Vitro
In 2006–2007, several groups of investigators reported the direct reprogramming in vitro of rodent embryonic and adult fibroblasts to pluripotent ESCs by retrovirally transfecting them with four transcription factor genes: Oct4, Sox2, Klf4, and c-Myc (Takahashi and Yamanaka,2006; Okita et al.,2007; Wernig et al.,2007; Blelloch et al.,2007; Meissner et al.,2007). This feat was subsequently repeated with human dermal fibroblasts using the same combination of genes (Takahashi et al.,2007; Lowry et al,2008) and a different set of transcription factor genes: Oct4, Sox2, Nanog, and Lin28 (Yu et al.,2007). The efficiency of reprogramming is low (less than 0.3%) and slow, taking 2 or more weeks. In line with what is known from studies on ESCs, Oct4 and Sox2 appear capable by themselves of conferring pluripotency, because fibroblasts could be reprogrammed even if Klf4, cMyc, Nanog, or Lin 28 were each left out of the mix (Takahashi et al.,2007; Yu et al.,2007; Nakagama et al.,2008; Wernig et al.,2008a). However, reprogramming was delayed even more and efficiency reduced even further in the absence of these transcription factors, indicating that they play a supporting role in the efficiency and speed of reprogramming (Jaenisch and Young,2008, for review). The low efficiency of reprogramming is due to a majority of cells being trapped in partially reprogrammed states characterized by incomplete chromatin modification and repression of differentiation-specifying transcription factors (Mikkelsen et al.,2008) that could be relieved by treatment with small molecules that inhibit chromatin-modifying enzymes. The efficiency of reprogramming was increased 3- to 5-fold over controls by treatment of cells with the DNA methyltransferase inhibitor 5-aza-cytidine or small interfering or short hairpin RNAs against Dnmt1 (Mikkelsen et al,2008), and 10-fold with 5-aza-cytidine plus dexamethasone (Huangfu et al.,2008). Three small inhibitors of deacetylases also increased the efficiency of reprogramming. Suberoylanilide hydroxamic acid increased efficiency by 2-fold, trichostatin by 10-fold, and valproic acid by over 100-fold (Huangfu et al.,2008).
The reprogrammed cells are called induced pluripotent stem cells (iPSCs; Takahashi et al.,2007). Comparative analysis of mouse and human iPS cells and ESCs revealed that the two were virtually identical in morphology, growth characteristics, global patterns of gene activity, and DNA methylation status of pluripotency gene promoters. Mouse and human iPS cells reactivated pluripotency genes and the silent X chromosome, and up-regulated telomerase (Maherali et al.,2007; Okita et al.,2007; Wernig et al.,2007; Takahashi et al.,2007; Yu et al.,2007; Park et al.,2008). The cells differentiated into representatives of all three germ layers after intramuscular injection into SCID mice, and mouse iPSCs formed germline-competent adult chimeras when injected into 2N or tetraploid blastocysts (Okita et al.,2007; Wernig et al.,2007; Hanna et al.,2008). The potential of these cells for regenerative medicine has been demonstrated in a humanized mouse model of sickle cell anemia, where hematopoietic cells derived by directed differentiation of iPS cells eliminated the disease (Hanna et al.,2007), and in a rat model of Parkinson's disease, where iPSC-derived dopaminergic neurons elicited functional improvement when transplanted into the brain (Wernig et al.,2008b). Furthermore, iPS cells have been generated from the skin fibroblasts of an 82-year old female patient with familial amyotrophic lateral sclerosis (ALS) and directed to differentiate into motor neurons (Dimos et al.,2008). Such cells will be useful in determining whether reprogramming and directed differentiation results in healthy functional cells and to study the interaction between genetic, epigenetic, and environmental factors in the development of disease. Other uses for iPS cells will be drug toxicity analysis, therapeutic drug discovery, and construction of bioartificial tissues and organs.
Mechanism of Reprogramming
The mechanism by which exogenously delivered transcription factors reprogram cultured somatic cells is not yet clear. While the end result may be the same, the path to pluripotency does not appear to be the same in iPS cells as that induced by egg cytoplasm after SCNT. A major difference is that reprogramming by SCNT is complete within 2–3 days, whereas fibroblast reprogramming by exogenous pluripotency gene constructs is gradual. Using doxycycline-induced retroviral constructs to transiently express Oct4, Sox2, c-myc, and Klf4 in mouse fibroblasts, and analyzing the expression of the fibroblast surface antigen Thy-1 and the ESC surface antigen SSEA-1, Stadtfeld et al. (2008) showed that reprogramming in iPS cells takes 10–13 days and involves the production of successive subpopulations of cells that become progressively more able to become iPS cells after withdrawal of doxycycline. Brambrink et al. (2008) obtained similar results. The endogenous pluripotency genes are activated late, in cells that strongly express SSEA-1 and lack Thy-1 expression; these cells are the ones that most frequently become bona fide iPS cells. These intermediate cell populations, which reflect discrete steps of reprogramming, will be useful in analyzing pluripotency gene activation, differentiation gene silencing, and changes in epigenetic marks and chromatin organization that confer iPSC status.
Biological and Bioethical Issues With iPS Cells
Although direct reprogramming of somatic cells by exogenous pluripotency genes bypasses immunorejection, and eliminates the bioethical issues surrounding SCNT-derived human ESCs, there will be several problems with iPSCs to be overcome on the way to the clinic, some the same as SCNT-derived ESCs (Cyranoski,2008). Chief among them is the necessity to have a 100% differentiation frequency to avoid the possibility of teratocarcinoma formation. Different cell lines would vary in this respect, and each line would have to be rigorously tested. The time required to produce iPSCs is estimated to be as long as 2 years, given the number of steps involved, from establishing and expanding the cell line, differentiating and expanding precursor cells, and testing the cells for teratocarcinoma or cancerous tumor formation. Thus, the cost of production is likely to be high, perhaps several times more expensive than patient-specific skin grafts, which run as much as $100,000 (Cyranoski,2008). Furthermore, different lines of human iPS cells, like human ESCs (Carpenter et al.,2004; Osafune et al.,2008), will exhibit highly variable HLA profiles and differentiation bias. These problems might be solved by establishing banks of multiple hiPSC lines with different HLA profiles and differentiation biases, along the lines of the registries, banks and information-sharing that have been established for human ESCs (O'Rourke et al.,2008). We will also need to insure safe and effective ways to deliver iPS cells to damaged target tissues, which in turn will require in-depth understanding of homing and migration signals and pathways (Laird et al.,2008; Mooney and Vandenburgh,2008, for reviews).
Lastly, there is the important question of how robust is the differentiation of iPS cells, and how they will behave over long periods of time. Is the extent of the genetic, biochemical and structural differentiation of iPSC derivatives the same as that of the same cell type produced by natural (in vivo) developmental processes? Will they maintain the level of differentiation required for function over the life of the individual? For example, would β-cells differentiated from iPS cells have the complete complement of metabolic machinery, glucose-sensing receptors and level of insulin output as normal β-cells? Will adult stem cells derived from iPS cells be able to self-renew as they repair an injury when transplanted into regenerating tissues, as observed for grafted satellite cells? How long do the differentiated progeny of iPS cells survive after transplantation? Do any of the cells revert to iPSC status with the possibility of teratoma formation? We can possibly answer some of these questions by making embryos chimeric for labeled iPS cells and following cell survival, structure, and function over the natural life span of the animal (which, to our knowledge, has never been done). Mice are the norm for making chimeras, but we must be cautious about generalizing results from such short-lived animals to long-lived primates such as humans, where cells must survive and function over much longer periods of time.
Human iPS cells will not be without their own ethical issues (Cyranoski,2008; Sugarman,2008). The relative ease of making iPSCs means that it could be done clandestinely for the wrong reasons. Theoretically, it will be possible to direct the differentiation of male iPSCs to oocytes and sperm (female cells are not able to differentiate into sperm, because this requires the Y chromosome) for the purpose of in vitro fertilization (the embryo would not be a clone, because of chromosomal exchange during the meiotic divisions). Furthermore, the cells could be genetically manipulated for superior quality. One could envision a black market for embryos derived from this type of “self-fertilization.” Controversial experiments, such as grafting human neural precursors derived from iPS cells into nonhuman primates (Greene et al.,2005), would be made easier and more difficult to monitor. Due to these concerns and others, the government of Japan has forbidden the “implantation of embryos made with iPS cells into human or animal wombs, the production of an individual in any other way from iPS cells, the introduction of iPS cells into an embryo or fetus, and the production of germ cells from iPS cells” (Cyranoski,2008).
Further Research on iPS Cells
To increase the simplicity, efficiency, and authenticity of somatic cell reprogramming, we need to know in more detail the molecular events and pathways that lead to pluripotency (cf., Mikkleson et al.,2008). In addition, it will be interesting and instructive to compare the global gene activity patterns and epigenetic characteristics of iPS cells with those of ASCs and multipotent or pluripotent cells recovered from fresh or cultured bone marrow or from connective tissue cultures (Jiang et al.,2002; Young and Black,2004; Kucia et al.,2007). Culture-derived multipotent cells have been reported to share some of the pluripotency gene expression and surface antigens of ESCs. Knowing how closely they mimic ESCs would help in understanding whether some somatic cells are easier to reprogram than others. iPS cells can be derived from adult somatic cells other than fibroblasts, such as hepatocytes, gastric epithelial cells (Aoi et al.,2008), and terminally differentiated B lymphocytes (Hanna et al.,2008). Unlike the iPS cells derived from fibroblasts, which require a large number of retroviral integration sites for reprogramming, hepatocyte and gastric epithelium-derived iPS cells appear to require only a few integration sites, and thus do not cause tumor formation in hosts (Aoi et al.,2008). Surprisingly, however, chimeric mice generated from these cells die postnatally, suggesting the existence of a subtle reprogramming defect. A recent significant finding is that neural progenitor cells (NPCs) can be reprogrammed to iPS cells using only two pluripotency transcription factor genes, Oct4 and Klf4, suggesting that ASCs may be easier to reprogram than differentiated somatic cells (Shi et al.,2008). The fact that NPCs already express endogenous Sox2 may be a major factor in the ability to reprogram them using only two other pluripotency genes.
A major step forward would be the ability to reprogram somatic cells in vitro without the introduction of accessory DNA. How might this be done? Deeper knowledge of the reprogramming mechanism will be vital, but we might start by introducing the Oct4 and Sox2 RNAs or proteins into permeabilized somatic cells. Because there undoubtedly are other as yet unknown factors involved in reprogramming, somatic cell or oocyte protein extract might be useful as a reprogramming agent (Fig. 5). Hakelien et al. (2002) reported the reprogramming of 293T fibroblasts in vitro by extract of Jurkat cells into cells that partially resembled T-lymphocytes and to cells with some characteristics of neurons using a neuronal cell precursor extract. Hansis et al. (2004) exposed human 293T kidney cells and primary leukocytes to Xenopus oocyte extract and detected up-regulation of the pluripotency markers alkaline phosphatase and Oct4. Two candidate reprogramming molecules in Xenopus oocyte extract have been identified, the chromatin remodeling ATPase, BRG1 (Hansis et al.,2004), and a tumor-associated protein, Tpt1 (Koziol et al.,2007). Antibody depletion of BRG1 from the extract abolished its reprogramming activity. Adding back BRG1 restored the activity. Antibody depletion of Tpt1 reduced both Oct4 and Nanog transcription by HeLa cell nuclei transferred into enucleated oocytes. If oocyte extract can induce reprogramming of somatic cells, it would be worth fractionating the extract to determine an active fraction, then doing a proteomic analysis of that fraction to narrow down the identity of the active proteins.
Another way to generate iPSCs would be to screen combinatorial chemical libraries of synthetic small molecules on somatic cells to identify candidates that will initiate the activity of the pluripotency transcriptional network (Fig. 5). Shi et al. (2008) have shown that the small molecule BIX-01294, which inhibits the G9a histone methyltransferase, improves the reprogramming efficiency in Oct4/Klf4-transduced neural progenitor cells to a level comparable to that achieved with a four-factor transduction, consistent with a function involving de-repression of these genes. A system of high throughput screening on adult somatic cells will undoubtedly reveal other molecules that have effects on reprogramming. Such a screening system has been devised for the maintenance of self-renewal of human ESCs, and several molecules have been identified, based on Oct4 and Nanog expression, that maintain short-term self- renewal in the absence of FGF-2 (Desbordes et al.,2008).
PERSPECTIVE ON THE FUTURE: TOWARD THE CHEMICAL INDUCTION OF REGENERATION IN VIVO
The discovery of how to induce pluripotency in adult somatic cells, made possible by research on ESCs, leads directly to the next step in the evolution of regenerative medicine: how to chemically (pharmaceutically) induce the regeneration of tissues in vivo from residual healthy cells at the site of the injury, using natural or artificial regeneration scaffolds, soluble molecules or a combination of both (Stocum,2006, for review). The goal of this section is to discuss research findings that indicate the feasibility of such a strategy, as well as to present general technical approaches and two specific comparative nonmammalian models of regeneration that have high value in identifying the molecular components for use in the chemical induction of regeneration. In addition, we will point out what we see as the limitations of chemical induction and cell transplantation strategies in regenerative medicine.
Chemical induction therapies would be significantly less invasive, expensive and time-consuming than the use of cell transplants. They would involve the suppression of scar-promoting molecules in nonregenerating tissues that harbor stem cells, the reprogramming of differentiated cells to become stem cells in nonregenerating tissues devoid of stem cells, or the transdifferentiation (reprogramming without dedifferentiation) of one cell type to another. Rather than reprogramming differentiated cells to ESCs, the objective would be to dedifferentiate them to adult stem cells, which would then recapitulate only that part of the developmental program necessary to restore the tissue. This reprogramming might involve some of the same epigenetic mechanisms that confer pluripotency (for example, Polycomb group proteins and chromatin remodeling enzymes), but would more likely require introducing (and/or activating) lineage-specific transcription factors, rather than the transcription factors that confer pluripotency on ESCs. For example, analysis of Oct4 and pseudogene expression suggests that Oct4 is not active in adult stem cells (Leidtke et al.,2007).
Feasibility of a Chemical Induction Approach In Situ
Zhou et al. (2008) have demonstrated the feasibility of in vivo transdifferentiation of one cell type to another by means of the introduction of lineage-specific transcription factor genes. They converted mouse pancreatic exocrine cells to β-cells by adenoviral transfection of the exocrine cells with a cocktail of genes for the transcription factors Ngn3, Pdx1, and Mafa, which are important regulators of pancreatic and β-cell differentiation. When the transfection was performed in mice rendered diabetic by streptozotocin, hyperglycemia was reversed. These results, if they can be repeated in nonhuman primates, have potentially significant implications for the treatment of human diabetes and other diseases that affect single cell types.
The feasibility of chemical induction of regeneration in situ without the introduction of exogenous DNA into cells is indicated by investigations aimed at developing therapeutic interventions to prevent scarring of skin in humans (Ferguson and O'Kane,2004, for review) and the ability of neural precursor cells of the CNS to differentiate into new neurons when removed from a regeneration-inhibitory environment (Kempermann,2005).
In adult humans, like other adult mammals, skin wounds heal by forming scars. During scarring, the missing normal tissue is replaced with an abnormally organized extracellular matrix consisting predominantly of fibronectin and collagen types I and III. By contrast, skin wounds applied to mammalian embryos during the first one-third to half of gestation heal with no signs of scarring. Surprisingly, the sterile aqueous environment of the amniotic fluid is irrelevant to the scar-free healing. Investigations of wound healing and scarring in the pouch young of marsupials have shown that, despite frequent contamination with maternal urine and feces, skin wounds of early pouch young heal with no scars (Armstrong and Ferguson,1995). Comparison of the cellular factors involved in embryonic healing without scars and adult healing with scars has revealed differential expression in isoforms of transforming growth factor beta (TGFβ). Embryonic wounds display low levels of TGFβ1 and TGFβ2, whereas adult wounds contain large quantities of these two isoforms (Whitby and Ferguson,1991; Cowin et al.,2001).
Experimental manipulation of skin wound healing has supported the hypothesis that the isoforms 1 and 2 of TGFβ are causally involved in the formation of scars. Application of neutralizing antibodies to TGFβ1 and/or TGFβ2 to healing wounds of adult rodents, at the time of wounding or soon thereafter, reduces scarring (Shah et al.,1992,1994,1995). On the basis of these observations, a formulation of mannose-6-phosphate, which inhibits the activation of TGFβ1 and 2, has been developed to improve or prevent scarring in the adult skin (see www.renovo.com). The efficacy of this agent, Juvidex, is currently being tested in phase-2 clinical trials.
A chemical induction of regeneration also appears feasible in the CNS. Prime targets for such a therapeutic approach are the two areas of the mammalian brain in which adult neurogenesis occurs continuously—the subventricular zone of the lateral ventricle and the subgranular zone of the dentate gyrus. However, the absence of constitutive neurogenesis in other parts of the brain does not necessarily reflect an intrinsic lack of cells with neurogenic potential in these regions. Instead, several lines of evidence demonstrate that it is the lack of appropriate signals in the microenvironment that prevents proliferation, differentiation, migration, and/or survival of neuronal progenitor cells in such regions, thus forcing them to remain quiescent in vivo (Emsley et al.,2005; Kempermann,2005, for reviews). Such cells have been isolated from various areas of the brain and the spinal cord (Reynolds and Weiss,1992; Weiss et al.,1996; Palmer et al.,1999; Kondo and Raff,2000). Their neurogenic potential has been revealed after treatment with appropriate exogenous growth factors in vitro, or after exposure of these cells to a permissive environment in vivo.
For example, proliferation of cells in the ependymal zone of the central canal increases dramatically after spinal cord injury, but in vivo these cells appear to give rise to astrocytes only (Johansson et al.,1999). However, when such cells are transplanted into the dentate gyrus of the hippocampus, they differentiate into granule cell neurons (Shihabuddin et al.,2000). Similarly, when precursor cells from the adult rat hippocampus are grafted into the rostral migratory pathway, they migrate into the olfactory bulb and differentiate into olfactory interneurons (Suhonen et al.,1996). By contrast, when these precursor cells are implanted into the cerebellum—a non-neurogenic site in the mammalian brain—their phenotype is restricted to glia. These experiments strongly support the notion that, after molecular modification of the local microenvironment, it may be possible to direct cell proliferation and subsequent development of endogenous neural precursors.
Genomic and Proteomic Approaches to Identifying Inducing Molecules
The challenge to the development of a chemical induction-based therapy without introduction of exogenous DNA is to identify the molecular signals that define a regeneration-permissive environment, and to mimic such an environment under therapeutic conditions. The traditional approach to the first problem has been to select a candidate gene or protein and characterize its expression pattern and functional role during tissue regeneration. The decision of what to select is often based on information available through developmental studies: cellular signals that regulate developmental processes during embryonic or adult development are frequently also involved in repair processes during tissue regeneration, and are thus good candidates for further examination.
An alternative to this “single candidate approach” has become available with the advent of genomics and proteomics. Both approaches enable investigators a more unbiased, large-scale profiling of regeneration-associated genes or proteins. Subtractive hybridization analysis has been used to analyze changes in regenerating Xenopus limbs at different time points after amputation and between regeneration-competent and deficient developmental stages (King et al.,2003; Grow et al.,2006). Differentially regulated genes have been identified by microarray analysis in fetal and neonatal wounds at different time points after wounding (Colwell et al.,2008). A large number of genes that govern homeostasis and regeneration in planaria have been identified by siRNA knockdown experiments (Reddien et al.,2005; Sanchez-Alvarado,2006). One of these, the transcriptional regulator β-catenin, acts as a molecular switch that specifies anteroposterior polarity during regeneration and homeostasis (Gurley et al.,2008; Petersen and Reddien,2008).
In stem cell biology, large-scale identification at the gene level is being augmented by global analysis at the protein level (see Whetton et al.,2008, for commentary). Proteomic analysis has become particularly powerful when studying protein abundance in tissue obtained under two different conditions, for example, intact state vs. regenerating state. This approach is referred to as differential proteomics. The advantage of a proteomic approach is that it detects only the actual effectors of gene activity, whereas genomic approaches detect transcripts that may or may not be translated into proteins. An authoritative discussion of the different methodologies available can be found in Monteoliva and Albar (2004).
Investigations using differential proteomics in mammals after traumatic brain injury have revealed two major classes of proteins (Zupanc,2007, for review): proteins that reflect the degenerative processes after injury; and proteins involved in the attempts of the mammalian brain to counteract the degenerative effects and to repair the damaged tissue. For example, 24 hours after closed head injury in early postnatal rats the abundance of β-actin, α-tubulin, and β-tubulin is significantly decreased, compared with sham-treated controls (Jenkins et al.,2002). This decrease reflects a degradation of these cytoskeletal proteins. Furthermore, the abundance of 78-kDa glucose-regulated protein precursor is reduced. This decrease of the precursor molecule is likely to result in a lowered abundance of 78-kDa glucose-regulated protein. The latter has been shown to have a neuroprotective function by reducing apoptotic cell death (Tsuchiya et al.,2003). Thus, the overall effect of the reduced abundance of the 78-kDa glucose-regulated protein precursor could be to increase the number of cells undergoing apoptotic cell death after injury. On the other hand, Cu/Zn superoxide dismutase displays an increase in protein abundance. This increase may reflect an attempt of the brain to restrict damage because experimental evidence suggests that this antioxidant enzyme protects tissue from cell death by reducing the production of free radicals (Reaume et al.,1996; Yin et al.,2001).
In addition to changes in the abundance of whole proteins, differential proteome analysis has also indicated the occurrence of proteolytic processes after injury. For example, 48 hours after closed head injury, cortical tissue collected from adult rats has revealed a breakdown of collapsin response mediator protein-2 (Kobeissy et al.,2006). In Caenorhabditis elegans, this molecule is required for proper axon morphology and pathfinding (Hedgecock et al.,1985; Siddiqui and Culotti,1991). Thus, the inactivation of collapsin response mediator protein-2 as a developmentally important molecule provides one explanation why, even under conditions in which new cells are generated after injury, these cells commonly fail to further develop in the adult mammalian brain (Arvidsson et al.,2002).
Regeneration of Complex Structures: Comparative Analysis of Regeneration-Competent and Deficient Tissues in Nonmammalian Models
A strategy to identify natural regeneration promoting and inhibiting signals for the chemical induction of regeneration of structures composed of multiple integrated tissues is to use a comparative analysis of regeneration-competent vs. regeneration-deficient systems or conditions. Information about the cellular factors mediating successful regeneration in a regeneration-competent system often provides clues about cellular factors that negatively interfere with the attempt to regenerate in a regeneration-deficient system, and vice versa. This has been done in rodent models with regard to regeneration of fetal vs. adult skin (Ferguson and O'Kane,2004, for review) and ear and heart tissue in wild-type vs. mutant MRL mice, which exhibit gain of regenerative capacity (Heber-Katz et al.,2004, for review; Caldwell et al.,2008). The identification of signals, natural or synthetic, that induce the activation and differentiation of stem cells, and/or the reprogramming of somatic cells in vivo, will also be facilitated by the study of animals that have much higher stem cell proliferative capacity than mammals, or have the innate ability to reprogram differentiated somatic cells to stem-like cells. Here, we discuss two such nonmammalian models, the CNS of the teleost fish and the limb of the salamander.
CNS of teleost fish.
The adult brain of the teleost fish Apteronotus leptorhynchus is one of the best-examined nonmammalian regeneration-competent systems (Zupanc,2006,2008; Zupanc and Zupanc,2006, for reviews). In the intact teleostean brain, the rate of continued cell proliferation, relative to the total number of brain cells, has been estimated to be at least one, if not two, orders of magnitude higher than in the intact mammalian brain (Zupanc and Horschke,1995; Zupanc and Zupanc,2006; Hinsch and Zupanc,2007). These new cells originate from pluripotent adult stem cells harbored in dozens of specific proliferation zones within the brain (Hinsch and Zupanc,2006). Approximately half of the new cells persist for the rest of the fish's life and develop into a variety of cell types, including neurons and glial cells (Zupanc et al.,1996,2005; Ott et al.,1997; Hinsch and Zupanc,2007).
This neurogenic potential in the intact brain is closely linked to an enormous regenerative ability of teleost fish. After mechanical lesions applied to the cerebellum, teleosts are capable of restricting cell loss to an initial wave of cell apoptotic death, thus lacking a secondary wave of cell death common in the mammalian brain (Zupanc et al.,1998). Furthermore, there is no evidence for necrosis, a common type of cell death after injuries in the mammalian cerebellum. In the injured teleostean brain, the cells lost to injury are replaced within a few weeks. As shown in Figure 6, these cells are recruited by inducing an increase in mitotic activity of stem cells in areas exhibiting constitutive neurogenesis, and by activating stem cells that are quiescent in the intact brain (Zupanc and Ott,1999).
Differential proteome analysis has revealed some of the proteins potentially mediating this regenerative potential (Zupanc et al.,2006). Three days after application of a mechanical lesion to the cerebellum, protein abundance is reduced or increased in about equal numbers of proteins. Twenty-four of these differentially regulated proteins have been identified. They include cytoskeletal proteins essential for the formation of new cells and proteins mediating the correct assembly of these structural proteins; proteins potentially involved in cell proliferation, cellular motility, neuroprotection, and energy metabolism; and a potential transcription regulator, bone marrow zinc finger 2.
The information obtained through proteome analysis of a regeneration-competent system becomes particularly useful when comparing it to the regulation of protein abundance after injury in a regeneration-deficient system, such as the adult mammalian brain. For example, the abundance of glutamine synthetase is increased 3 days after cerebellar lesions in fish (Zupanc et al.,2006), whereas this enzyme is down-regulated under traumatic conditions in the mammalian central nervous system (Grosche et al.,1995; Härtig et al.,1995; Lewis et al.,1989,1994; Oliver et al.,1990; Smith et al.,1991). Glutamine synthetase is a glia-specific enzyme that converts synaptically released glutamate into the nontoxic amino acid glutamine. Under normal conditions, this mechanism prevents the accumulation of neurotoxic amounts of glutamate in neural tissue and thus protects neurons from cell death. However, under traumatic conditions the amount of glutamine synthetase is insufficient to catalyze the excessive amounts of glutamate released by damaged cells. The additional down-regulation of glutamine synthetase after brain trauma in mammals aggravates the situation in mammals, thus failing to limit the spread of damage caused by continuous overexcitation of postsynaptic glutamate receptors. On the other hand, the up-regulation of glutamine synthetase in the teleostean brain is likely to enable fish to reduce the neurodegenerative effect induced by glutamate neurotoxicity. Although largely unexplored, such differences between regeneration-competent species and regeneration-incompetent species, revealed by comparative proteome analysis, are likely to prove extremely powerful to identify target mechanisms for therapeutic intervention after traumatic brain injury in humans.
Amphibian limb regeneration.
The limbs of larval and adult urodeles are histologically and morphologically complex structures that regenerate after amputation by the formation of a blastema at the amputation surface. The blastema is composed of mesenchymal stem-like cells derived from both adult stem cells (satellite cells of muscle) and by the dedifferentiation of mature skeletal, muscle, dermal and Schwann sheath cells (Morrison et al.,2006; Stocum,2004,2006, for reviews). Dedifferentiation involves histolysis of the tissues by ECM-degrading matrix metalloproteinases (MMPs) and acid hydrolases, cellularization of muscle, loss of phenotypic specialities, and re-entry into the cell cycle (Brockes and Kumar,2005, for review). Growth of the blastema is regulated by Fgfs from the wound epidermis and nerves, and the newt anterior gradient protein (nAG) produced first by the Schwann sheath cells and then by the wound epidermis (Kumar et al.,2007). As the blastema grows, it is patterned into the missing limb parts. Growth and patterning of the blastema recapitulate the molecular, histological and morphological changes observed in limb development (Endo et al.,2004). Patterning is regulated by a variety of Hox genes, Meis2, and signals such as Wnts and Shh (Gardiner and Bryant, 1995,1996; Mercader et al.,2005). What is so unique about the salamander limb is its ability to form a blastema after amputation by reprogramming the nuclei of differentiated limb cells to an adult stem cell-like state. If we can understand this reprogramming, there is the potential to apply it to mammalian cells in situ (Odelberg,2002; Straube and Tanaka,2006).
Currently, little is known about the molecular regulation of limb cell dedifferentiation. Numerous genes involved in salamander and newt limb regeneration have been identified (Geraudie and Ferretti,1998, for review; King et al.,2003; Grow et al.,2006), some of which are up-regulated and others down-regulated in the initiation and maintenance of de-differentiation. Up-regulated genes include the transcription factor msx1, Nrad, radical fringe, and notch. Nrad is the orthologue of a diabetes-associated mammalian ras protein, and is up-regulated in the nuclei of newt limb myofibers within 4 hours after amputation (Shimizu-Nishikawa et al.,2001). Radical fringe and notch are ligand and receptor, respectively for the Notch pathway that maintains self-renewal of mammalian stem cells, and are expressed in blastema cells (Cadinouche et al.,1999). The downstream targets of these proteins in regenerating limbs are unknown. Newt myofibers appear to have a receptor, not shared by mammalian myofibers, that responds to a thrombin-activated protein in vertebrate serum by re-entering the cell cycle (Tanaka et al.,1997,1999). This ligand and its receptor have not been identified. Nevertheless, regenerating newt limb protein extract induces C2C12 myofibers to cellularize and dedifferentiate in vitro to mesenchymal stem cells capable of becoming osteoblasts and adipocytes when exposed to differentiation factors (McGann et al.,2001; Fig. 7). Little is known about the role played by chromatin modifying enzymes, Polycomb group proteins, or micro RNAs in the reprogramming of differentiated urodele limb cells to stem cell status during regeneration. However, Yakushiji et al. (2007) have reported hypomethylation of the Shh enhancer in the intact and regenerating limb of the axolotl, and moderate methylation of this enhancer in the intact and regenerating limbs of the newt, suggesting that low levels of methylation are associated with the function of this patterning gene.
In contrast to urodele limbs, anuran hindlimbs regenerate only at early tadpole stages, when the limb buds are still undifferentiated (Suzuki et al.,2006, for review). The capacity for regeneration is lost in the differentiated regions as they develop in a proximal to distal direction. By late tadpole stages, amputated hindlimbs regenerate only a spike of cartilage lacking any trace of asymmetric pattern, or nothing at all. This transition from regeneration-competence to deficiency is associated with a transition from hypomethylation to hypermethylation of the Shh enhancer (Yakushiji et al.,2007).
The contrast between the regeneration-competent urodele limb and early tadpole hindlimb vs. the regeneration-deficient late tadpole/froglet hindlimb offers the opportunity for comparative analyses that will reveal the differences in molecular components and networks that lead to the production of adult-type stem cells by dedifferentiation in one case, but not the other. Once known, the mammalian counterparts of molecules associated with regeneration competence could be used to chemically induce regeneration in regeneration-deficient limbs directly at the amputation site. Differences in gene activity between regeneration-competent and deficient hindlimb buds of Xenopus have been partly revealed by subtractive hybridization of cDNA libraries (King et al.,2003). A comparative proteomic analysis between early and late tadpole hindlimbs of Xenopus or between hindlimbs of the axolotl (Ambystoma mexicanum) and Xenopus froglets at different time points after amputation would be even more instructive.
At the same time, assays for specific chromatin modifying enzymes, examination of histone methylation and acetylation differences, and analysis and interference with micro RNAs might begin to reveal details of the epigenetic changes that characterize dedifferentiation in regenerating limbs and how these may differ in regeneration-deficient anuran limbs. Another strategy would be to analyze the proteomic and epigenetic differences of axolotl wild-type vs “short toe” limbs. “Short toes” is a loss of function mutant in which the digits of the limbs have fewer and shorter phalangeal elements than normal and which regenerate even more hypomorphically (Sato and Chernoff,2007, for review).
We can also screen combinatorial chemical libraries on amphibian tissues or whole limbs in vitro to identify small synthetic molecules that might mimic the natural dedifferentiation induction system. Several such molecules have already been identified that induce osteogenesis by MSCs (puromorphine), promote the self-renewal of mouse ESCs (SC1) and neuronal differentiation of multipotent hippocampal neural progenitor cells (neuropathiazol) and convert C2C12 myoblasts to neurons (neurodazine; Ding and Schultz,2004; Xu et al.,2008, for reviews). Another molecule, a tri-substituted purine called myoseverin, causes cellularization of C2C12 mouse and newt myofibers (Rosania et al.,2000). Myoseverin disrupts microtubules and up-regulates growth factor, immunomodulatory, ECM-remodeling and stress-response genes, consistent with the activation of pathways involved in wound healing and regeneration, but does not activate a full program of myogenic dedifferentiation (Duckmanton et al.,2005). Still another molecule, reversine, is a di-substituted purine that induces multipotency in mouse myoblasts and osteoblasts, and mouse and human fibroblasts (Chen et al.,2004). This process involves growth arrest by activation of the PI3K pathway and enhancement of Polycomb gene expression, suggesting that the mechanisms of differentiation gene silencing in ESCs and ASCs may be partly similar. Reversine-treated mouse dermal fibroblasts have been shown to participate in muscle regeneration when injected into the injured tibialis anterior muscle (Anastasia et al.,2006). Finally, retinoic acid is a small natural molecule involved in many aspects of cell differentiation. When administered in excess to amputated limbs of salamanders, it causes excessive de-differentiation and reprogramming of positional identity to more proximal, posterior and ventral values (Crawford and Stocum,1988a,b; Maden,1997; Stocum,2006, for reviews). How these molecules affect the transcriptional program of adult somatic cells in amphibians or mammals is a question well worth pursuing.
Limits of Regenerative Therapies
The technologies of regenerative medicine can be applied straightforwardly in cases of tissue injury due to trauma or ischemia. Their application to tissues damaged by degenerative diseases, however, will depend on the nature of the disease. Diseases caused by cell-autonomous (genetic) defects, such as hematopoietic genetic diseases, osteogenesis imperfecta, and muscular dystrophy, will be curable by wild-type cell replacement. Chemical induction of regeneration would not be of benefit for these diseases because regenerated tissues would come from cells that harbor the original defect. Autoimmune diseases such as type 1 diabetes, and rheumatoid arthritis, or diseases such as osteoarthritis, amyotrophic lateral sclerosis, Parkinson's disease, and Alzheimer's disease, which appear to be noncell autonomous, may not gain long-term benefit from regeneration either by cell transplant or chemical induction, because the extrinsic factors that cause the disease will remain. For example, Parkinson's disease is an ongoing process that appears to attack transplanted cells in a manner similar to host dopamine neurons in the substantia nigra (Kordower et al.,2008). Curing the diseases by eliminating the conditions that cause and maintain them would allow the tissues to then be regenerated by chemical induction. Chemical induction will be of definite benefit for the regeneration of tissues damaged by ischemia and/or trauma, such as spinal cord or brain injury, injuries to normal musculoskeletal tissues, and myocardial infarction.
The need to cure many diseases to gain the benefit of regenerative therapy highlights the great potential to understand their genesis by studying ES or iPS cells derived from patients who suffer from these diseases. What are the pivotal events and interactions that turn cells to the dark side? Having understood these, it may be possible to devise individualized therapies to eliminate the disease process, and then regenerate the tissue.
We thank Ursula M. Wellbrock for preparing the figures, and Marianne Zupanc, Nandini Rao, Fengyu Song, and Bingbing Li for critiques of the manuscript. During the writing of this review, D.L.S was supported as a Fellow of the International Center for Transdisciplinary Studies of Jacobs University Bremen.