Chapter 6: Muscle

## **Myogenic Precursors from iPS Cells for Skeletal Muscle Cell Replacement Therapy**

#### **Isart Roca, Jordi Requena, Michael J. Edel and Ana Belén Alvarez-Palomo**

**Abstract:** The use of adult myogenic stem cells as a cell therapy for skeletal muscle regeneration has been attempted for decades, with only moderate success. Myogenic progenitors (MP) made from induced pluripotent stem cells (iPSCs) are promising candidates for stem cell therapy to regenerate skeletal muscle since they allow allogenic transplantation, can be produced in large quantities, and, as compared to adult myoblasts, present more embryonic-like features and more proliferative capacity *in vitro*, which indicates a potential for more self-renewal and regenerative capacity *in vivo*. Different approaches have been described to make myogenic progenitors either by gene overexpression or by directed differentiation through culture conditions, and several myopathies have already been modeled using iPSC-MP. However, even though results in animal models have shown improvement from previous work with isolated adult myoblasts, major challenges regarding host response have to be addressed and clinically relevant transplantation protocols are lacking. Despite these challenges we are closer than we think to bringing iPSC-MP towards clinical use for treating human muscle disease and sporting injuries.

Reprinted from *J. Clin. Med*. Cite as: Roca, I.; Requena, J.; Edel, M.J.; Alvarez-Palomo, A.B. Myogenic Precursors from iPS Cells for Skeletal Muscle Cell Replacement Therapy. *J. Clin. Med.* **2015**, *4*, 243–259.

#### **1. Introduction**

Skeletal muscle is a dynamic organ in which an efficient regeneration process ensures repair after damage. The process of muscle regeneration creates new myofibers after necrosis resulting from injury or a degenerative process. The myonuclei of multinucleated myofibers are post mitotic, arrested in the G0 phase of the cell cycle and unable to proliferate. A resident population of adult myogenic stem cells called "satellite cells" is the main player in the regeneration process. These cells reside in a quiescent state, located between the basal membrane and the plasmalemma of each myofiber. Upon signaling from the damaged myofibers, satellite cells become activated, undergo an asymmetric division to self-renew, and produce activated myoblasts that are able to proliferate, migrate to the site of injury, and fuse with the existing myofibers or to form new myotubes [1]. Besides satellite cells, other populations with stem cell properties have been described as capable of undergoing myogenesis and contribute to myofiber repair, such as mesangioblasts, bone marrow-derived stem cells, pericytes, or interstitial muscle-derived stem cells, though it appears that *in vivo* they contribute to a much smaller extent than satellite cells [2].

Repeated cycles of myofiber necrosis and regeneration in muscle dystrophies (MD), such as Duchenne muscular dystrophy (DMD) and some limb girdle dystrophies, result in exhaustion of satellite cell regenerative capacity in humans [3]. Similarly, neuromuscular diseases in which neuromuscular junctions are lost and muscles undergo subsequent atrophy, such as spinal muscle atrophy (SMA) and familiar amyotrophic lateral sclerosis (ALS), present deficiencies in the satellite cells compartment [4,5]. Moreover, the myofibers in both MDs and neuromuscular diseases present different abnormalities in their structure and functionality [6–8]. Other situations in which muscle regeneration is compromised are severe injury [9] and inflammatory myopathies [3]. Restoration of the satellite cell compartment with healthy cells would restore the regenerative capacity of the muscle and progressively substitute the defective myofibers. Therefore, in all of these conditions, myogenic cell replacement therapy provides a promising perspective for the treatment of degenerative myopathies.

#### **2. Using Myoblasts as a Cell Therapy**

Transplantation of donor myoblast or satellite cells isolated from healthy individuals has been tried extensively in the past with somewhat positive but insufficient results and scarce references to functional improvement [10]. In 1995, allogenic normal myoblasts were transferred into the biceps brachii arm muscles of DMD patients in order to restore the lack of dystrophin protein [11]. Although some fusion of donor nuclei into host myofibers was observed, there was no significant improvement in muscle function. Genetic correction has also been explored to allow for autologous transplantation of expanded myoblasts, but results again showed engraftment but a low contribution to host fibers [12]. Massive death of most of the transplanted cells within a few days after intramuscular delivery has been reported by several laboratories [13]. The reasons why the myoblasts die initially are not clear but probably relate to immune aspects, anoikis, and a hostile environment in the host damaged muscle. Moreover, using myoblasts as a donor source poses a limitation in the amount of original tissue for cell isolation from normal human muscle biopsies. It also limits the possibilities of *in vitro* expansion because myoblasts are limited to a few passages due to senescence and the decreased self-renewal capacity of the cells due to the expansion process [14]. Therefore, it is difficult to obtain a clinically relevant number of transplantable myoblasts from a donor source. The use of other adult stem cells, with high proliferative capacity, as an alternative source of myogenic cells has been investigated with disappointing or inconclusive results such as bone marrow-derived stem cells [15], pericytes [16], and mesangioblasts [17]. Further research is needed to establish the efficacy of cell therapy using these types of donor cells.

Clinical trials using myogenic cell therapy to treat muscular dystrophies started in the 1990s, showed some engraftment of the donor cells but no clear signals of disease recovery or symptom alleviation (see Table 1).

However, extensive preclinical and clinical work over the past few decades has helped to identify some relevant issues to address in order to improve cell therapy in muscular dystrophies. The main limitations of this therapy are transplanted cell engraftment and contribution to host myofibers, which seems to be highly dependent on survival—immunosuppression is thus required but other factors might be contributing as well—and migration out of the site of injection. The transplantation regime can also affect engraftment success [18].

Taking all this into account, the ideal donor cell for skeletal muscle regeneration should be easily accessible and able to expand extensively without losing myogenic and engraftment capacity, have a great survival and fusion rate with host myofibers (high myogenic capacity), and be highly motile to spread within the muscle. Moreover, it should contribute to the satellite cell compartment, enabling indefinite muscle regenerative capacity. Finally, the ideal myogenic donor cell should have low immunogenicity, and be able to be delivered systemically, since intramuscular injection does not seem a feasible approach given the large volume of muscle tissue to be treated.

However, extensive preclinical and clinical work over the past few decades has helped to identify some relevant issues to address in order to improve cell therapy in muscular dystrophies. The main limitations of this therapy are transplanted cell engraftment and contribution to host myofibers, which seems to be highly dependent on survival—immunosuppression is thus required but other factors might be contributing as well—and migration out of the site of injection. The transplantation regime can also affect engraftment success [18].

Taking all this into account, the ideal donor cell for skeletal muscle regeneration should be easily accessible and able to expand extensively without losing myogenic and engraftment capacity, have a great survival and fusion rate with host myofibers (high myogenic capacity), and be highly motile to spread within the muscle. Moreover, it should contribute to the satellite cell compartment, enabling indefinite muscle regenerative capacity. Finally, the ideal myogenic donor cell should have low immunogenicity, and be able to be delivered systemically, since intramuscular injection does not seem a feasible approach given the large volume of muscle tissue to be treated.



#### **3. Induced Pluripotent Stem Cells (iPSCs)-Derived Myogenic Progenitors (iPSC-MP)**

Embryonic stem cells (ESC) are pluripotent stem cells derived from the inner cell mass of a blastocyst that are able to self-renew and to be differentiated in all tissues in the body. Induced PSCs share most of the features of ESCs but are derived from adult somatic cells, e.g., dermal fibroblasts, by the transient expression of a defined set of reprogramming factors [28]. The fact that iPSCs do not involve the destruction of embryos, with the consequent ethical issues, and allow for autologous production of the pluripotent cells has opened up an enormous range of possibilities for the regenerative cell therapy field. Since iPSCs have limitless replicative capacity *in vitro* and can differentiate into myoblast-like cells, they represent an attractive source of myogenic donors for muscle regeneration. Induced PSC-MP also represents a highly valuable tool for *in vitro* drug testing and disease modeling for muscular genetic conditions that were so far limited because of the difficulties of obtaining large quantities of tissue.

Initially, human ESCs (hESCs) proved to be difficult to differentiate into myogenic progenitors, probably due to the fact that paraxial mesoderm and subsequently the myogenic program are not well recapitulated during embryoid body (EB)—three-dimensional aggregates of pluripotent stem cells—formation [29]. The first protocols using different sequential culture conditions, including a mesenchymal differentiation step, were successful at producing myogenic progenitors capable of engrafting *in vivo* but these protocols were lengthy and inefficient [30]. It has been reported that the need for a mesodermal transition previous to a myogenic commitment is determined by the epigenetic landscape in human ESCs [31]. Higher efficiency and shorter protocols were designed by overexpression of myogenic transcription factors. Pax3 and Pax7 are paired box transcription factors that contribute to early striated muscle development and are expressed in the dermatomyotome of paraxial mesoderm. Darabi and colleagues showed that inducible expression of Pax3 using viral vectors at early EB formation overcame mesoderm patterning restrictions and yielded up to 50% myogenic cells within barely a week [29]. Albini *et al.* described how overexpression of MyoD1—a transcription factor that appears after Pax3 and Pax7 in muscle development and in activated satellite cells—alone could not induce myogenic commitment directly on hESCs, but concomitant overexpression of the chromatin remodeling complex component BAF60C overcame the mesodermal transition limitation [32]. In opposition to these results, Rao *et al.* describe hESC-derived myogenic progenitors by inducible lentiviral overexpression of MyoD1 directly on hESC cells, without a previous EB formation [33].

Other more efficient and genetic modification-free protocols have been described to obtain myogenic progenitors from hESCs, such as isolation of the PDGFRĮ<sup>+</sup> population from EB derived-paraxial mesoderm [34] or isolation of the SM/C-2.6+—satellite cell-like—population from differentiating mouse ESC-derived EB cultured in high serum [35].

Since the appearance of iPSCs, extensive work has been done to obtain myogenic progenitors with a vision to their clinical application and disease modeling (Table 2). The first iPSC-MP came from mouse cells using a protocol similar to the one described above for ESC [35], based on spontaneous differentiation and sorting of SM/C2.6 positive cells [36]. Similarly, the group of Awaya reported a method of deriving mesenchymal cells with myogenic capacity from EB by a protocol based on selective enrichment though step-wise culture conditions [37]. The resulting cells showed long-term engraftment in immunocompromised mice pre-injured with cardiotoxin, and evidence of replenishing the satellite cell compartment. However, these protocols are long and not very efficient. Using an inducible lentiviral expression system, Darabi *et al.* produced satellite cell-like progenitors by overexpression of Pax7—a transcription factor required for somite myogenesis in the embryo and a marker for satellite cells in the adult—in EB from mice (miPSCs) and humans (hiPSCs) [38,39]. The resulting cells were able to engraft in a mouse model of muscular dystrophy and to produce regeneration and restore some muscle strength, and even showed evidence of donor-derived satellite cells—by expression of Pax7 and M-cadherin by the capacity of regeneration after a subsequent injury. They reported much better proliferative capacity of the myogenic progenitors *in vitro* and much better engraftment as compared to myoblasts. Lentiviral inducible overexpression of Pax3 in iPSCs from dystrophin-lacking mice, which were gene corrected with a truncated version of dystrophin (ȝ-utrophin), produced in a similar fashion myogenic progenitors that engrafted, differentiated, and repopulated the satellite cell compartment and exhibited neuromuscular synapses [40]. Goudenege and colleagues described a two-step protocol consisting of first culturing in a myogenic medium and then infecting with an adenovirus expressing MyoD1 that rendered myogenic progenitors able to engraft in the muscular dystrophy model mdx mice [41]. Also, using a self-contained, drug-inducible expression vector, based on the PiggyBac transposon for overexpression of MyoD1 and an efficient and quick conversion of undifferentiated iPSCs into myogenic progenitors with the ability to engraft in immunocompromised mice has been described [42]. A limitation on the use of MyoD1 for generating myogenic progenitors is the induction of cell cycle arrest when expressed too long at high levels; therefore, as an excellent proliferative capacity is needed to expand *in vitro* and survive *in vivo*, careful dosage and timing are necessary when using this transcription factor.

Though gene overexpression approaches are fast, efficient, and appropriate to generate myogenic precursors for disease modeling, the risk of undesired genetic recombination or reactivation makes them unsuitable for a future application in the clinic for regenerative cell therapy. Different ways to obtain transplantable myogenic progenitors that do not involve any genetic modification and are still efficient and fast have recently been described. Recently, several reports describe other protocols without gene overexpression that include high concentrations of bFGF and EGF on free floating spheres [32] and, faster and more efficient, the use of GSK3 inhibitors and bFGF [43,44] in one of the cases, producing myogenic progenitors that engrafted in immunocompromised mice that contributed to the satellite cell pool [43].


**Table 2.** Protocols for myogenic progenitor derivation from iPSC and *in vivo* testing.


**Table 2.** *Cont.* 

\* NA = not assessed; \*\* Duchenne's Muscular Dystrophy; & Becker's Muscular Dystrophy; Ref.: Reference.

Another way of avoiding introducing exogenous DNA is the transfection of *in vitro*-synthesized mRNA to overexpress the required transcription factors for myogenic conversion. It was recently shown as a proof of principle that transfection of MyoD1 mRNA in hiPSCs produced myogenic cells with the ability to fully differentiate [45] *in vitro*.

Other cells with myogenic potential that are not myoblasts have been derived from iPSCs: the group of Tedesco has developed mesangioblast (pericyte progenitors)-like cells that have been tested in animal models [46].

#### **4. Disease Modeling**

The different approaches published so far to make myogenic progenitors from hiPSCs are good models of myogenesis *in vitro*, as the produced cells recapitulate the expression of markers observed *in vivo*. They are able to fuse to produce premature myofibers in the animal *in vitro* and in most cases they have been tested in animal models for engrafting and fusion with host fiber. Several reports describe the establishment of myogenic cell lines produced from iPSCs from patients with different types of muscular dystrophy. Human iPSC-MPs have been established using MyoD1 overexpression by a PiggyBac vector on hiPSCs: Miyoshi Myopathy, a distal myopathy caused by mutations in DYSFERLIN, patients' fibroblasts [42], and carnitine palmitoyltransferase II deficiency, is an inherited disorder that leads to rhabdomyolysis [47]. Duchenne muscular dystrophy, the most common type of MD, is due to a mutation in the dystrophin gene and has been modeled by adenoviral expression of MyoD1 [41] and by inducible lentiviral Pax3 overexpression [40]. The group of Hosoyama have also described the derivation of myogenic derivatives using their sphere-base culture system from hiPSCs from Becker's muscular dystrophy, spinal muscular atrophy, and amyotrophic lateral atrophy [42]. The created cell lines make great tools for drug screening and further research into the molecular mechanisms of the different myopathies, and can be obtained in large quantities with minimal patient invasion.

#### **5. Future Challenges for Clinical Application**

Myogenic progenitors made from iPSCs seem to be a promising candidate for stem cell therapy to regenerate skeletal muscle since they can be produced in large quantities and present more embryonic-like features, so are probably more motile and proliferative compared to adult myoblasts. However, even though results in animal models show an improvement from previous work with isolated myoblasts, in terms of fiber contribution and functional recovery [39,41], a clinically relevant transplantation protocol still needs to be designed.

#### *5.1. In Vivo Survival, Engraftment and Migration*

One of the major caveats of myoblast therapy was the massive death after transplantation. The inflammatory and immunological response to allogenic transplants probably played a role in the survival of the cells and also engraftment, migration, and differentiation [48]. However, myoblast death is seen before the onset of the immunological response and in the presence of immunosuppressors or for autologous transplantation, where there should be no immune response [21,23]. Also, anoikis and the toxic environment from the high oxidant stress that characterizes dystrophic muscles may play a role in the survival of cells. These challenges to survival will be encountered by hiPSCs-MP in the same ways as purified adult myoblasts. Regarding engraftment, all the published work on hiPSCs-MP in animal models shows *in vivo* engraftment and fusion with host cells, but greater extent is needed for a clinically relevant cell therapy protocol. Limited migration from the injection site, in part due to high mortality, but also to intrinsic capacity, is another major limitation that iPSC-derived cells must overcome to outperform myoblast therapy. Some authors describe iPSC-MP as resembling embryonic more than adult myoblasts [31]. The use of two markers expressed during embryogenesis by hypaxial migratory myogenic precursors, C-MET and CXCR4, has been proposed to isolate the most migratory fraction of hiPSC-MD [49]. Also, beta 1 integrin, expressed in satellite cells, is essential for engraftment [11] and can be another migratory phenotype selection marker.

#### *5.2. Fibrosis*

Another major limitation to regeneration is dense fibrotic tissue. TGF-ȕ1 induces collagen I deposition from myogenic cells with subsequent fibrotic tissue formation. Fibrosis limits myoblast engraftment as well as motility and this prevents axons from arriving to myofibers. Unfortunately, there are no drugs on the market that can overcome fibrosis in MD patients. However, there is a report that bone marrow-derived stromal cell transplantation in the muscle of an ischemia model reduced fibrosis due to paracrine effects [50]. This inhibitory effect should be studied in hiPSCs-MP if they are to be a candidate for use in a clinical setting.

#### *5.3. Creating the Perfect Niche*

Tissue engineering can also be of great help for the survival of transplanted myogenic progenitors in the hostile environment of a damaged tissue. Creating a three-dimensional niche for the transplanted myogenic progenitors that resembles satellite cells' natural niche *in vivo* by using biomaterials (alginate, collagen, and hyaluran) will conserve the engrafted cells' homeostasis and allow asymmetric division and myogenic commitment [51]. The cells to be transplanted would be seeded in the 3D scaffold and a graft generated *in vitro*. To complete the niche, extracellular matrix components and signaling molecules to stimulate proliferation, migration, and angiogenesis should be included. Muscle flaps made with decellularized devices from large mammals and synthetic scaffolds complemented with an *in vitro*-produced extracellular matrix from cell cultures derived from the host provide suitable tools for translation to the clinic [52]. From the complex set of requirements for skeletal muscle tissue engineered implants to function and integrate *in vivo*, some issues have already been addressed, such as restoration of the muscular-tendon junction or vascularization, while others like reinnervation still need further work [49].

#### *5.4. Genetic Correction vs. Immunocompatible Transplantation*

When addressing genetic origin myopathies, the transplanted cells should contain the correct version of the gene. This can be achieved in two ways: by genetic correction of patient-derived cells or by allogenic transplantation of immunocompatible donor cells. One of the major features of iPSCs is the possibility of generating patient-derived tissues with minor invasion. Several groups have performed gene correction on patient iPSCs. iPSC-derived mesangioblasts, from a Limb-Girdle MD patient, in which the wild-type alpha-sarcoglycan gene had been restored by lentiviral delivery, engrafted, and fused with host fibers when transplanted in nude mice [46]. Lamin A/C (LMNA) has also been corrected in laminopathy patient-derived iPSCs using a helper-dependent adenoviral vector, which is safer than other viral vector approaches [53]. Duchenne MD iPSCs have also been corrected with ȝ-utrophin using a sleeping beauty transposon system [39]. In any case, gene therapy is still under development and a totally safe way of gene correction has still not been demonstrated.

Another approach is to transplant cells created from a healthy donor that are matched for the main antigens in the host immunological rejection, the HLA antigens. An HLA-typed bank of iPSCs could be created to provide a source of compatible donor cells for the individual patients. A relatively small number of donors can provide an acceptable match to a high percentage of the population [54]. This approach would also be more feasible as a therapeutic approach than the expensive and time-consuming generation of personalized iPSC-MP.

It is necessary to take into account that in the case of genetic diseases that lack the native protein, its expression from the grafted tissue will most likely induce a considerable immune response that needs to be carefully addressed.

#### *5.5. Delivery Route*

Moreover, the desirable myogenic progenitor should be able to cross the blood barrier to allow for systemic delivery. Treatment of local damage could be done by local intramuscular injections or bio-engineered grafts, but for a cell therapy for MD, SMA, and ALS, in which all muscles in the body are affected, a systemic delivery is necessary. Very few reports show successful engraftment after intra-arterial delivery [38,39,46]. The adequate dosage and regime of injections still needs further study.

#### *5.6. Safety*

For all the reported work in humans and animals models using muscle stem cells, neither adverse side effect has been described, nor colonization in other organs when systemically delivered [39]. Also, for iPSC-MP no teratoma formation has been detected [37,39]. However, the double reprogramming process—first to pluripotency and then to myogenic lineage—bring along the risk of chromosomal abnormalities and genetic instability [55]. Darabi *et al.* described how, from several clones tested for *in vivo* engraftment and fiber contribution, those that performed better were the ones with a normal karyotype [38]. In this sense, chromosomal, genetic, and epigenetic studies must be performed on the cells to be transplanted before taking them to the clinic application. Also, reprogramming and differentiation methods should not include exogenous DNA but use, for example, mRNA transfection; the use of the oncogene c-Myc should be avoided when reprogramming for clinical applications. Genes involved in epigenetic remodeling [56] and cell cycle regulation [57] have been proposed as alternatives to c-Myc in reprogramming. In this regard, variants of c-Myc with no oncogenic potential such as L-Myc or the W136E c-Myc mutant are also able to induce reprogramming to pluripotency with less tumorigenic potential [58].

#### *5.7. Clinical Grade Protocols*

Whatever the method of choice is for generating the myogenic progenitors, a clinical grade protocol must be designed for the cells to be used in patients. The generation process should not include any viral vector or exogenous DNA, should be free of animal products, and should use as far as possible defined media to increase reproducibility and comply with good manufacturing procedures. Such a protocol has not yet been described for either iPSC generation or the derivation of MP.

#### **6. Conclusions**

The use of hiPSCs as a source of myogenic progenitors for cell therapy for the treatment of muscle degenerative diseases overcomes several of the limitations encountered in adult myoblast therapy: (i) easy non-invasive source of donor cells; (ii) unlimited proliferative capacity *in vitro*, and (iii) better performance when tested in mouse models *in vivo*—possibly because of more embryonic-like features. In recent years, several protocols of derivation of myogenic progenitors from iPSCs have been described reaching very satisfactory efficiency in a short time. The use of transcription factors (Pax7, MyoD1) overexpression or GSK3ȕ inhibitors has contributed greatly in this direction. However, a clinical grade protocol still needs to be described, including the definition of safety and genetic stability requirements for clinical applications. Also, isolation of the MP presenting the most promising features for successful regeneration *in vivo* could improve the performance of the cell therapy, such as selecting cells that are more migratory and proliferative or with the possibility of systemic delivery. Other limitations relating to the host—for example, the inflammatory and immune response and the appearance of fibrotic tissue—present a major hurdle to a cell therapy approach. More research with selective inhibitors or modulators of these processes is needed, and the use of bioengineering to create a 3D protective niche for the transplanted cells would contribute to the long-term success of a muscle stem cell therapy strategy.

#### **Acknowledgments**

Ana Belén Alvarez Palomo is supported by project grant BFU2011-26596 and FBG307900. Michael J. Edel is supported by the Program Ramon y Cajal (RYC-2010-06512) and project grant BFU2011-26596. We thank Miranda D. Grounds and Jovita Mezquita for critical readings of the manuscript.

#### **Author Contributions**

Isart Roca performed literature searches and bibliography compilations, contributed to prepare the tables, and to the writing of the manuscript. Jordi Requena contributed to literature searches and bibliography. Michael J. Edel contributed to draft, edit and revise the manuscript. Ana Belén Alvarez Palomo performed literature searches and bibliography compilations, drafted, formatted and wrote the manuscript, prepared the tables, and edited and revised the final manuscript.

#### **Conflicts of Interest**

The authors declare no conflict of interest.

#### **References**



Chapter 7: Bone

### **The Use of Patient-Specific Induced Pluripotent Stem Cells (iPSCs) to Identify Osteoclast Defects in Rare Genetic Bone Disorders**

#### **I-Ping Chen**

**Abstract:** More than 500 rare genetic bone disorders have been described, but for many of them only limited treatment options are available. Challenges for studying these bone diseases come from a lack of suitable animal models and unavailability of skeletal tissues for studies. Effectors for skeletal abnormalities of bone disorders may be abnormal bone formation directed by osteoblasts or anomalous bone resorption by osteoclasts, or both. Patient-specific induced pluripotent stem cells (iPSCs) can be generated from somatic cells of various tissue sources and in theory can be differentiated into any desired cell type. However, successful differentiation of hiPSCs into functional bone cells is still a challenge. Our group focuses on the use of human iPSCs (hiPSCs) to identify osteoclast defects in craniometaphyseal dysplasia. In this review, we describe the impact of stem cell technology on research for better treatment of such disorders, the generation of hiPSCs from patients with rare genetic bone disorders and current protocols for differentiating hiPSCs into osteoclasts.

Reprinted from *J*. *Clin*. *Med*. Cite as: Chen, I.-P. The Use of Patient-Specific Induced Pluripotent Stem Cells (iPSCs) to Identify Osteoclast Defects in Rare Genetic Bone Disorders. *J. Clin. Med.* **2014**, *3*, 1490–1510.

#### **1. Introduction**

Studying rare genetic bone disorders is clinically highly relevant. Although individual diseases only affect a small percentage of the population (less than 200,000 people or about 1 in 1500 people in the United States), overall, a large number of people suffer from these skeletal disorders due to their frequency (almost 500 rare genetic bone disorders listed by NIH Office of Rare Disease Research). Many of these diseases become apparent early in life and are present throughout the patient's entire life. The diverse expressivities of clinical manifestations, from lethality of newborns to mild skeletal abnormalities, make the diagnosis of some of these disorders challenging. Moreover, most of these rare bone diseases are understudied due to the rarity of human specimens and unavailable animal models, and therefore treatment options are often limited or lacking. It is thus important to establish better models for studying such disorders.

Research focusing on genetic disorders of the skeleton is not only beneficial for future treatment of patients, but has significantly contributed to our knowledge on key concepts of bone biology. Rare genetic bone disorders have been linked to abnormal bone development and/or bone remodeling. Pathologically and embryologically these diseases can be subdivided into four major groups: (1) disorders affecting skeletal patterning; (2) disorders of condensation/differentiation of skeletal precursor structures; (3) disorders affecting growth and (4) disorders of bone homeostasis caused by perturbation of interaction between the bone forming osteoblasts and the bone resorbing osteoclasts [1]. Our group has been studying a rare genetic bone disorder, craniometaphyseal dysplasia (CMD) characterized by progressive thickening of craniofacial bones and widening of metaphyses in long bones, utilizing a knock-in mouse model [2]. We have identified defects of osteoblasts and osteoclasts in mice with a CMD mutation [3]. We currently study CMD in a human system using patient-specific induced pluripotent stem cells (hiPSCs) to identify osteoclast defects and we believe this strategy can be applied widely for studying other rare genetic disorders.

In this article, we review some of the rare genetic bone disorders with osteoclast defects, the generation of hiPSCs from patients with rare genetic bone disorders and the protocols for differentiating hiPSC into osteoclasts.

#### **2. Rare Genetic Bone Disorders with Osteoclast Defects**

Osteoclasts are cells responsible for resorbing bone and work in close concert with osteoblasts to model the skeleton during growth/development and to remodel the bone throughout life. Osteoclasts are derived from the monocyte/macrophage lineage of hematopoietic stem cells and are multinucleated giant cells expressing marker genes, such as tartrate-resistant acid phosphatase (Trap), Cathepsin K, calcitonin receptor, nuclear factor of activated T-cells, cytoplasmic 1 (Nfatc1). Development of osteoclasts (osteoclastogenesis) involves (1) the commitment of hematopoietic stem cells into osteoclast precursors; (2) the fusion of mononucleated osteoclast precursors into multinucleated osteoclast syncytia; (3) the differentiation/maturation to functional osteoclasts. During stage 1, transcription factors PU.1, MITF and c-FOS are important determinants of lineage specification [4–6]. In addition, M-CSF signaling is necessary for proliferation and survival of osteoclast progenitors as first became obvious by the osteopetrotic phenotype of mice lacking the *M-CSF* gene (op/op mice) [7]. Interaction of RANKL, a member of TNF family and strongly expressed by osteoblasts, with its receptor RANK on osteoclast progenitors is necessary for the fusion of osteoclast precursors. The activation of RANK/RANKL signaling initiates a cascade of gene expression, including the expression of chemokines such as Mcp-1 to attract RANK+ mononuclear osteoclast precursors and molecules important for osteoclast fusion such as Atp6v0d2 and DC-Stamp [8,9]*.* The final step in differentiation to mature and functional osteoclasts involves the formation of a ruffled border and sealing zone. Lack of functional osteoclasts by disruption of these processes or failure of polarization and cytokine organization can lead to osteopetrosis [10].

Many rare genetic bone disorders are partially or primarily caused by osteoclast defects. Dysfunctional osteoclasts can result in too much bone while increased bone resorption can lead to decreased bone mass. Some diseases present a combination of osteosclerosis with osteolytic lesions. Studies on these disorders highlight the important roles of some specific proteins or signaling pathways during osteoclastogenesis.

#### *2.1. Diseases of Decreased Osteoclast Resorption*

Osteopetrosis: Three main forms of hereditary osteopetrosis are autosomal recessive osteopetrosis (ARO), intermediate autosomal recessive osteopetrosis (IARO) and adult dominant osteopetrosis (ADO), the most severe form being ARO. Mutations causing these diseases generally lead to lack of acid secretion in osteoclasts. ARO presents in infants with severe sclerosis of bone, an increased rate of fracture, extreme reduction of bone marrow space, hepatosplenomegaly, anemia, compression of cranial nerves and growth failure [11]. Infantile malignant osteopetrosis is generally lethal and can only be treated by early bone-marrow transplantation. IARO is a recessive form of osteopetrosis with renal tubular acidosis. The affected protein, carbonic anhydrase type II is highly abundant in osteoclasts, cerebral neurons and in renal intercalated cells. Clinical features of IARO patients include the milder form of osteopetrosis, mental retardation due to cerebral calcification and renal dysfunction [12]. Two distinct types of ADO are known [13]. ADO is characterized by a generalized diffuse osteosclerosis with the most pronounced thickening at the cranial vault in its type I form and the most pronounced abnormalities in vertebrae in type II (Albers Schonberg disease).

Pycnodysostosis: This is an autosomal recessive disorder and can be diagnosed during early infancy. The phenotype is milder than ARO with short stature, recurrent bone fracture, skull deformity and hypoplasia of facial bones, sinuses and clavicles. Long bones are hyperostotic with narrow medullary canals. The calvarium and base of the skull are sclerotic. The genetic defect for pycnodysostosis has been identified in Cathepsin K, a lysosomal cysteine protease required to degrade collagen in resorption lacunae of osteoclasts [14].

#### *2.2. Diseases of Increased Osteoclast Resorption*

Paget's disease of bone (PDB): PDB is a late onset bone disease starting at mid-life or later. Genetic predisposition together with environmental risks and other risk factors such as trauma or surgery contribute to its etiology. PDB affects single or multiple locations of the skeleton where focal bone resorption occurs and bone is replaced with soft, fibrous expansile tissue that result in characteristic enlarged and softened bone tissue. Clinical symptoms include bone pain, bone deformity, deafness, pathological fractures and osteoarthritis [15,16]. Although increased osteoclast activity is primarily the cause of PDB, excessive osteoblast activity reflected in elevated serum alkaline phosphatase has been reported [17], which could be a sign of increased bone turnover.

Juvenile PDB (JPDB): Different from Paget's disease of bone, JPDB usually occurs in infancy or early childhood, characterized by massive thickening of calvaria, widened diaphyses, and deformities of extremities and vertebrae [18]. Autosomal recessive mutations in *TNFRSF11B* result in a less efficient form of osteoprotegerin (OPG) with reduced affinity for RANKL or in a failure to express OPG protein. OPG is a decoy receptor for RANKL, thus regulating osteoclast formation. As a consequence, increased bone resorption coupled with increased bone formation, are seen in JPDB [19,20].

Familial expansile osteolysis (FEO): FEO is an autosomal dominant rare bone disorder characterized by osteolytic lesions in major bones of the appendicular skeleton during early adulthood. It can also result in deafness and premature tooth loss due to abnormalities in the middle ear and jaw [21,22]. Mutations identified in *TNFRSF11A* cause enhanced RANK-mediated nuclear factor-țB (NF-țB) signaling and increased bone remodeling [23].

Expansile skeletal hyperphosphatasia (ESH): ESH is characterized by expanding hyperostotic long bones, early onset deafness, premature tooth loss, episodic hypercalcemia and increased alkaline phosphatase activity; the skull and appendicular skeleton display hyperostosis and/or osteosclerosis [24].

Mutations responsible for these rare disorders affecting osteoclast activity are summarized in Table 1. Research of disorders mentioned above would be greatly enhanced if elaborate models for studying osteoclastogenesis and osteoclast function would be available. Induced pluripotent stem cells (iPSCs) could be one worthwhile avenue to study such disorders in humans. iPS cell approaches have been published for certain bone remodeling disorders described below.


**Table 1.** Mutations in rare genetic bone disorders with osteoclast defects.

ARO: autosomal recessive osteopetrosis; IARO: intermediate autosomal recessive osteopetrosis; ADOI: adult dominant osteopetrosis, type I; ADOII: adult dominant osteopetrosis type II; PDB: Paget's disease of bone; JPDB: Juvenile Paget's disease of bone; FEO: familial expansile osteolysis; ESH: expansile skeletal hyperphosphatasia; OMIM: Online Mendelian Inheritance in Man; GL: Grey lethal; N/A: not available.

#### **3. Generation of hiPSCs from Rare Genetic Bone Disorders**

hiPSCs, similar to human embryonic stem cells (hESCs), have the ability to self-renew indefinitely and in theory differentiate to any cell type when induced under appropriate conditions. The advances in hiPSCs technology opened new opportunities for medical research in disease modeling, drug screening, gene therapy and genome editing [33–35]. Generation of hiPSCs can, for example, be achieved by introduction of reprogramming factors, *OCT3/4*, *SOX2*, *KLF4* and *c-MYC* or *OCT3/4*, *SOX2*, *NANOG*, and *LIN28* [36–38]. Methods successfully delivering these reprogramming factors into somatic cells include transduction with retroviruses, lentiviruses, adenoviruses, piggyBac transposons, episomal vectors, RNA, or protein [37,39–44]. Many types of somatic cells have been reprogrammed into hiPSCs, including fibroblasts, keratinocytes, peripheral blood mononuclear cells, cord blood cells, T cells, dental pulp stem cells, dermal papilla cells from hair follicles and urinary cells [37,45–49]. Patient-specific hiPSCs provide unique opportunities for researchers to dissect the pathogeneses and identify potential treatment strategies for the rare genetic bone disorders by providing a virtually unlimited source of cells carrying the disease-causing mutations. hiPSCs can be differentiated into functional cells of interest in the skeletal system, including osteoclasts. hiPSC disease modeling has been established for several non-skeletal disorders including type I and type II diabetes, muscular dystrophy, amyotrophic lateral sclerosis, Parkinson's disease, glioblastoma, familial platelet disorder with predisposition to acute myeloid leukemia (FPD) [36,50–55]. There have been only few attempts to establish hiPSCs from patients with rare genetic bone disorders (see also Table 2).


**Table 2.** iPSCs generated from patients with rare genetic bone disorders.

OI: osteogenesis imperfecta; CMD: craniometaphyseal dysplasia; FOP: Fibrodysplasia ossificans progressiva; MFS: Marfan syndrome; MSC: mesenchymal stem cells.

Osteogenesis Imperfecta (OI): OI, also known as brittle bone disease, is characterized by brittle bones that are prone to fracture and are caused by mutations in the *COL1A1* or *COL1A2* genes in the majority of cases. Misfolded collagen overwhelms the protein degradation machinery of cells and leads to abnormal bone matrix deposition by osteoblasts. 8 Types of OI have been identified. There is currently no cure for OI and treatment focuses on the prevention of fractures and the maintenance of mobility [60]. Deyle *et al.* established mesenchymal cell cultures from discarded bone fragments of OI patients undergoing surgery and further inactivated mutant collagen genes by adeno-associated virus (AAV)-mediated gene targeting, thus preventing the expression of misfolded collagen protein. OI and gene-targeted OI mesenchymal cells have been reprogrammed to hiPSCs [56].

Craniometaphyseal dysplasia (CMD): CMD is characterized by progressive hyperostosis of craniofacial bones and widened metaphyses of long bones. Patients often suffer from blindness, deafness, facial paralysis and severe headache due to hyperostosis and compression of the brain and nerves. Mutations for the autosomal dominant form of CMD have been identified in of progressive ankylosis (*ANKH*) gene and for a recessive form in Connexin 43 (*Cx43*) [61–63]. Our group identified dysfunctional osteoclasts in knock-in (KI) mice carrying a Phe377del mutation and in human osteoclast cultures [2,3]. The increased bone mass phenotype in CMD mice (*Ank*KI/KI mice) can partially be rescued by bone marrow transplantation. We have established a simple and efficient method to generate integration-free hiPSCs from peripheral blood of CMD patients and healthy controls using the Sendai virus, a cytoplasmic RNA viral vector [57], that can easily be removed from cells after reprogramming to iPSCs.

Fibrodysplasia ossificans progressiva (FOP): FOP is a rare genetic disorder caused by hyperactive mutations in the bone morphogenetic protein (BMP) type I receptor ACVR1 [64]. It is characterized by progressive ossification of soft tissues. The mechanism of heterotopic ossification is endochondral bone formation which involves pre-cartilaginous, fibro-proliferative and mineralization stages. Matsumoto *et al.* generated hiPSCs from skin fibroblasts of FOP patients and controls and showed increased *in vitro* chondrogenic differentiation and mineralization in FOP hiPSCs compared to wild type hiPSCs [58].

Marfan syndrome (MFS): MFS is a life-threatening, autosomal dominant disease with mutations identified in *FIBRILLIN-1* (*FBN1*) [65]. It is a disorder of fibrous connective tissue involving three systems: skeletal, cardiovascular and ocular. Skeletal features include long limbs and digits, deformities of vertebrae (scoliosis, thoracic lordosis) and anterior chest, increased height, and mild to moderate joint laxity. Quarto *et al.* generated hiPSCs from MFS patients and studied the pathogenic skeletogenesis *in vitro* [59]. They show that MFS-hiPSC faithfully represent the impaired osteogenic differentiation as a consequence of activation of TGF-ȕ signaling and revealed a crosstalk between BMP and TGF-ȕ signaling in MFS [66].

#### **4. Differentiating hiPSCs into Osteoclasts**

#### *4.1. Differentiating Mouse Embryonic Stem Cells (mESCs) into Osteoclasts*

Several studies have reported the generation of osteoclasts from mESC lines by culturing mESCs directly on a culture plate or by co-culturing mESCs with mouse bone marrow-derived stromal cells (ST2) or with the newborn calvaria-derived stromal cell line (OP9) from mice deficient in macrophage colony-stimulating factor (M-CSF) or through EB formation (for details see Table 3). Information gained from these mouse ESCs/iPSCs studies provided the fundamentals for establishing methods to generate osteoclasts from human ESCs and iPSCs.



#### *4.2. Commitment of Human ESCs/hiPSCs into Hematopoietic Lineages/OC Precursors*

Consistent and adequate hematopoietic differentiation of hiPSCs is a prerequisite step for differentiating hiPSCs into osteoclasts. Hematopoiesis during embryogenesis starts with the formation of the primitive streak, mesoderm differentiation and hematopoietic specification. *In vitro* studies have shown that hiPSCs can be differentiated into different hematopoietic lineages through similar processes. Inducing hematopoiesis from human embryonic stem cells (hESCs) or hiPSCs has been reported by several extensive studies using three systems: (1) differentiation by coculturing hESC/hiPSCs with stromal cells; (2) differentiation through formation of embryoid bodies (EB), which can be differentiated into the three germ layers including mesoderm; (3) differentiation by monolayer culture of hESCs/hiPSCs on extracellular matrix protein coated-plates, such as collagen IV. We summarize culture conditions of these protocols in Table 4.

There are advantages and disadvantages inherent to these protocols. Differentiation efficiency of co-culture methods relies largely on optimized cell densities of hESCs/hiPSCs and the mouse stromal cell lines. It can be challenging to have both cell culture systems ready for co-culture at the same time. On the other hand, co-culture method requires less hematopoietic cytokines and is therefore relatively inexpensive compared to protocols involving EB and monolayer cultures. Concentrations of cytokines used are critical in those cultures. Stimulatory or inhibitory effects of cytokines towards hematopoiesis can be observed depending on the concentrations used. It is difficult to directly compare the lengths of cell culture time or hematopoietic differentiation efficiencies among the approaches described in Table 4. Variability of culture conditions used for the experiments among these protocols is discussed below.

#### *4.3. Marker Genes for Mesodermal Formation and Hematopoietic Differentiation*

Marker genes for mesodermal and hematopoietic lineages have been used to determine the efficiency of hematopoietic differentiation protocols during *in vitro* differentiation. Temporal expression patterns of certain marker genes are reliable indicators of mesoderm and hematopoietic differentiation from undifferentiated hESCs or hiPSCs. Early mesoderm formation is indicated by the expression of marker genes such as *Brachyury* (*T*), *Mix paired-like homeobox* (*Mixl1*), *Goosecoid homeobox* (*GSC*), and silencing of the pluripotency genes [74,75]. While kinase insert domain receptor (a type III receptor tyrosine kinase, *KDR*), also known as endothelial growth factor receptor 2 (*VEGFR-2*) or fetal liver kinase 1 (*Flk1*), is already detected in hESC/hiPSCs, its expression increases during the transition from mesoderm to hematopoietic lineage [76]. In addition, transcription factors stem cell leukemia (*Scl/Tal-1*), runt-related transcription factor 1 (*Runx1*), globin transcription factor 1 (*GATA-1*) and *GATA-2* play important roles in hematopoietic commitment during embryogenesis [77–79]. Combinations of surface marker expression are used to detect the hematopoietic stem cells (HSC) or the maturation of HSCs into specific hematopoietic cells. CD34, CD31 and VE-cadherin are expressed in early hematopoietic cells and vascular associated tissues [76]. CD45 is a pan-leukocyte marker [80]. Lin<sup>í</sup>CD34+CD43+CD45+ cells represent a population of enriched myeloid progenitors [81].




#### *4.4. Factors and Cytokines to Promote Hematopoiesis*

Defined factors/cytokines can be added to support hematopoietic cell proliferation or differentiation under controlled conditions. BMP4 alone can induce primitive streak and early hematopoietic gene expression, including *Mixl1*, *Brachyury*, *Goosecoid*, *KDR*, *Runx1* and *Gata2* while BMP4 together with VEGF can increase expression of *Scl* and *CD34* [91]. The addition of FGF2 during hematopoietic differentiation of hESCs increases total cell number by improving cell proliferation but not cell survival [91]. A mixture of cytokines including stem cell factor (SCF), fms-like tyrosine kinase receptor-3 ligand (Flt-3), interleukin-3 (IL-3), interleukin-6 (IL-6) and granulocyte colony-stimulating factor (G-CSF) and thrombopoietin (TPO) is commonly used in hematopoietic differentiation protocols for hESCs. These cytokines have been shown to play important roles in maintenance of human hematopoietic cells [92,93]. Addition of SCF in the presence of BMP4, VEGF and FGF can significantly increase the yields of hematopoietic progenitors and mature cells *in vitro* [91]. When added to IL-3 and GM-CSF, SCF has profound effects on *in vitro* proliferation of primitive hematopoietic progenitors [94].

#### *4.5. Variability among Hematopoietic Differentiation Protocols*

Some studies showed variable efficiencies for deriving hematopoietic cell populations from hESCs/hiPSCs. Many factors contribute to this variability including the somatic cell type used for hiPSC reprogramming; the method of deriving hiPSCs; incomplete removal of reprogramming transgenes in hiPSCs; the culture conditions for maintaining hiPSCs; the type of differentiating medium, growth factors and hematopoietic cytokines added in the differentiation protocol; the dosage of cytokines to promote hematopoiesis; the oxygen level of cultures (normal oxygen or hypoxia); and the size of EBs or the densities of stromal cell lines [95]. It has been suggested that epigenetic memory exists in hiPSCs and the differentiation phenotype may be influenced by their cells of origin [96]. Transgenes remaining in hiPSCs can have a negative impact on hiPSC differentiation [97]. Multiple concentrations of BMP4 (5, 10, 25, 50 ng/mL) were tested and showed no differences in promoting CD34+ CD45+ populations [85] while Pick *et al.* showed BMP4 increases primitive streak and hematopoietic gene expressions in a dose-dependent manner [91]. It is therefore important to develop a method that eliminates as many variable factors as possible and that aims to obtain hematopoietic cells from hiPSCs by a reproducible and efficient protocol.

#### *4.6. Differentiating hiPSCs-Derived Osteoclast Progenitors into Osteoclasts*

Two publications describe the successful differentiation of hiPSCs to hematopoietic cell stage and further to mature OC progenitors and into functional osteoclasts. Choi *et al.* cultured a Lin<sup>í</sup>CD34+CD43+Cd45+ population in the presence of GM-CSF and vitamin D3 on poly (2-hydroxyethyl methacrylate)-coated plates for 2 days to expand osteoclast progenitors and then induced osteoclast maturation in cultures with Į-MEM, 10% FBS, and MTG solution supplemented with GM-CSF (50 ng/mL), Vitamin D3 (200 nM) and RANKL (10 ng/mL) [81]. Grigoriadis *et al.* cultured myeloid precursors derived from hiPSCs in IMDM containing 10% FCS, M-CSF (10 ng/mL) and RANKL (10 ng/mL). Osteoclasts were defined as multinucleated cells (3 nuclei) by TRAP positive staining and the capability of resorbing bone/dentin chips. Expressions of OC marker genes such as Cathepsin K, calcitonin receptor, NFATc1 are increased in these functional OCs [87].

Studies summarized in Tables 3 and 4 demonstrate that osteoclasts can be generated from both, human and mouse ES/iPSCs. In general, differentiation of osteoclasts from hESCs/hiPSCs systems is more challenging than differentiation from mouse ESCs/iPSCs. Differentiation of hESCs/hiPSCs requires more hematopoietic cytokines, longer culture periods and generally results in less efficient osteoclast formation. However, the mechanisms behind differences between human and mouse osteoclastogenesis are still unclear.

#### *4.7. Strategies of Using hiPSC-Osteoclasts to Study Rare Genetic Bone Diseases*

hiPSC technology in general enables researchers to reprogram somatic cells into an ES-like state followed by differentiation into desired cell type. While human osteoclasts can be differentiated directly from peripheral blood, the big advantage of hiPSC technology is that osteoclasts can be generated without repeated sampling of patients. hiPSCs provide a virtually unlimited cell source to study molecular mechanisms of osteoclastogenesis with the potential to develop therapies, which is especially important when studying rare genetic bone diseases. Because of potential species-specific differences, studying abnormal osteoclastogenesis in the human system may be closer to clinical reality than using animal models.

A preferred strategy for studying defective osteoclastogenesis in rare genetic bone disorders using patient-specific hiPSCs is summarized in Figure 1. One challenge of hiPSC disease modeling *in vitro* is the lack of genetically matched controls. Using healthy subjects as controls may not be the best solution as individual hESCs and hiPSC lines differentiate to specific cell populations with variable efficiency because of biological variability [98]. Distinguishing mutation-relevant disease phenotypes from genetic/epigenetic variations becomes easier in isogenic hiPSCs, which only differ at disease-causing mutations. Correction or introduction of specific mutations into a cell can be achieved by genome editing using zinc-finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs) or the clustered regulatory interspaced short palindromic repeats/Cas9 system (CRISPR) [99–102]. Using step-wise OC differentiation protocols for differentiating hESCs or hiPSCs may allow researchers to identify which step of osteoclastogenesis is disrupted by a disease causing mutation. Analysis tools are available to study each step (lineage determination of precursors, precursor proliferation, fusion to multinucleated syncytia, maturation to functional osteoclasts) such as expression of marker genes, numbers of TRAP+ mono/multinucleated cells, resorption efficiency, live-image migration assays and nuclear localization of NFATc1. Therapeutic strategies can be investigated once the pathologic mechanisms are understood.

#### **Figure 1.** Summary of generating osteoclasts from human iPSCs.

#### **5. Conclusions**

Osteoclast defects are involved in many rare genetic bone diseases as well as in some common bone diseases such as osteoporosis. Although human OC can be cultured from peripheral blood, being able to differentiate OCs from hiPSCs has at least two additional advantages: (1) eliminating the need to repeatedly obtain blood from study subjects; (2) serving as an *in vitro* model for studying hematopoiesis during embryogenesis. Lessons learned from embryology and differentiation studies are expected to improve protocols for consistent and efficient differentiation of hiPSCs into hematopoietic cells and further into osteoclasts. Similar concepts can be applied to differentiate hiPSCs to other bone cells such as osteoblasts. We believe this model will have great impact on a better understanding of bone diseases and to establish the bases for potential therapies.

#### **Acknowledgments**

This work is supported by National Institutes of Health funding (NIDCR, K99/R00 DE021442).

#### **Conflicts of Interest**

The author declares no conflict of interest.

#### **References**



Chapter 8: Germ Cells

### **Human iPS Cell-Derived Germ Cells: Current Status and Clinical Potential**

#### **Tetsuya Ishii**

**Abstract:** Recently, fertile spermatozoa and oocytes were generated from mouse induced pluripotent (iPS) cells using a combined *in vitro* and *in vivo* induction system. With regard to germ cell induction from human iPS cells, progress has been made particularly in the male germline, demonstrating *in vitro* generation of haploid, round spermatids. Although iPS-derived germ cells are expected to be developed to yield a form of assisted reproductive technology (ART) that can address unmet reproductive needs, genetic and/or epigenetic instabilities abound in iPS cell generation and germ cell induction. In addition, there is still room to improve the induction protocol in the female germline. However, rapid advances in stem cell research are likely to make such obstacles surmountable, potentially translating induced germ cells into the clinical setting in the immediate future. This review examines the current status of the induction of germ cells from human iPS cells and discusses the clinical potential, as well as future directions.

Reprinted from *J. Clin. Med*. Cite as: Ishii, T. Human iPS Cell-Derived Germ Cells: Current Status and Clinical Potential. *J. Clin. Med.* **2014**, *3*, 1064–1083.

#### **1. Introduction**

There are various reasons to generate germ cells from human pluripotent stem cells in the laboratory. First, *in vitro* recapitulation of gametogenesis and early embryogenesis using such induced germ cells is expected to enhance our understanding of the basis of human reproduction because the inaccessibility to human eggs (oocytes) and embryos has hampered relevant research. Second, human germ cell induction research will establish a precious platform for modeling infertility and congenital anomalies that have been difficult to study using animals. Third, the *in vitro* induction of germ cells from autologous pluripotent stem cells should lead to a new form of assisted reproductive technology (ART) for infertile patients who wish to have genetically-related children.

Recent advances in stem cell research have made it conceivable that human sperm (spermatozoon) and oocytes will be induced from pluripotent stem cells in the near future. Notably, a Japanese group reported that mouse embryonic stem (ES) cells and induced pluripotent (iPS) cells could be differentiated into fertile spermatozoa and oocytes via primordial germ cell (PGC)—like cells, and demonstrated that viable offspring could be derived from pluripotent stem cells [1,2]. Although their protocols used gonadal tissues and an *in vivo* induction system, their work established an important step on the path to the *in vitro* recapitulation of gametogenesis. Significant progress has also been made in the differentiation from both human ES cells [3–8] and iPS cells [8–13] into human germ cells over the last decade. A recent report demonstrated that human iPS cells can be indirectly or directly differentiated into the male germline, including haploid, round spermatid-like cells [10,12,13]. Rapid advances in stem cell research would help to overcome the current technical issues and lead to the *in vitro* formation of bona fide human spermatozoa and oocytes.

If functional oocytes and spermatozoa can be differentiated from human iPS cells, the use of such cells for research will contribute to the molecular elucidation of gametogenesis, as well as the onset and progression of various diseases in obstetrics, gynecology, and neonatology/pediatrics. However, with regard to the reproductive use of such germ cells induced from autologous iPS cells, sufficient preclinical research will need to be performed to confirm the safety of the offspring. Remarkably, the overview of ART (Appendix) using induced germ cells appears to occur against the Weismann barrier, wherein hereditary information moves only from germ cells to somatic cells [14]. Such germ cells are likely to be subject to genetic and/or epigenetic instabilities during iPS cell generation and germ cell induction. Moreover, although assessing the biological function of induced germ cells involves the creation of embryos and subsequent culture for a short period, human embryo research is strictly regulated in most countries [15]. In this review article, the current status of germ cell induction from human iPS cells is examined and discussed in light of clinical potential and future directions.

#### **2. Clinical Implications of Germ Cell Induction** *in Vitro*

Two fundamental cell types constitute multicellular eukaryotes. Somatic cells proliferate by mitosis and form the tissues and organs comprising the body. Germ cells undergo meiosis as well as mitosis, resulting in the generation of gametes that can transfer half the genetic material to the next generation. The lineage of germ cells is referred to as the germline.

If germ cells can be efficiently induced from human iPS cells, the availability of such germ cells could contribute to various biomedical fields. First of all, the research use of human female germ cells and embryos is largely difficult owing to ethical reasons and the scarcity of oocytes and embryos for research. In contrast, patient-specific induced germ cells can model diseases that are derived from aberrant germ cells or that occur during embryogenesis. A wide variety of somatic cells which are differentiated from patient-specific iPS cells have already been used for disease-modeling to enhance the understanding of the pathogenesis of diseases [16]. Currently, the low efficiency of the differentiation of human iPS cells into germ cells has hampered the unveiling of the molecular pathogenesis of various diseases, including germ cell tumors [17], aneuploidy, sex chromosome abnormalities [11], and female and male infertilities.

If functional germ cells are induced from iPS cells, such germ cells are also expected to impact ART treatment (Figure 1). Although ART has helped many infertile patients to produce offspring, the current ART procedures are based on the premise that an infertile couple can produce fertile gametes in order to perform intrauterine insemination (IUI), *in vitro* fertilization (IVF), or intracytoplasmic sperm injection (ICSI) (Appendix). Otherwise, the couple must use donor gametes. This option has raised ethical issues and social confusion. ART using donor gametes results in the birth of genetically-unrelated children. Such children born of donor gametes frequently confront stigma that stems from being uninformed about their genetic parents or due to their lack of resemblance to their parents in shape and appearance [18]. In addition, some sperm donors have anonymously provided their gametes to a tremendous number of patients, creating social problems [19]. Such cases frequently occur because there are many prospective parents who have no viable gametes due to congenital anomalies, or because they have been rendered sterile by receiving chemotherapy and radiation therapy for cancer treatment [20–22], or because the females have undergone age-related oocyte senescence [23].

Recent progress in germ cell induction research is increasing the possibility of a new form of ART using germ cells induced from autologous iPS cells for patients with no viable gametes (Figure 1). If fertile spermatozoa can be induced from a male patient's iPS cells, performing IVF or ICSI will be possible using the generated spermatozoa. Similar approaches can be performed when fertile oocytes are generated from iPS cells. Even if no mature spermatozoa are obtained from the induction, *in vivo* spermatogenesis could be restored by transplanting spermatogonial stem cells (SSCs) derived from autologous iPS cells into the testis of a male patient [24–26]. In 1997, infusions of oocyte cytoplasm including mitochondria from donor oocytes was conducted in order to enhance the fertility of quality-compromised oocyte with mitochondrial defects [23], resulting in the birth of over 30 children [27]. However, the U.S. Food and Drug Administration concluded that further research was required for the use of this procedure in humans due to potential health risks to the progeny [28]. If this ooplasmic transfer procedure is sufficiently improved and induced female germ cells which genetically match the patient's oocytes can be obtained from iPS cells, such germ cells could be used as a resource for ooplasmic transfer. Following such ART procedures, the resulting embryos can be carefully examined for three to five days post-conception, and one or more viable embryo(s) can then be selected for embryo transfer. Thus, autologous iPS-derived germ cells are expected to meet the reproductive needs of infertile couples who have lost viable gametes for medical reasons or aging but wish to have genetically-related children.

**Figure 1.** The potential reproductive uses of iPS cell-based germ cells. Autologous iPS cells can be generated from somatic cells biopsied from infertile patients who have lost viable oocytes or spermatozoa. Subsequently, germ cells are induced from the iPS cells. The regenerated germ cells can be used for *in vitro* fertilization or intracytoplasmic sperm injection to create embryos for transfer. In cases of male infertility, spermatogonial stem cells (SSCs) could be transplanted into patients to restore spermatogenesis potential. In cases of female infertility, ooplasmic transfer to enhance the viability of quality-compromised oocytes is conceivable if female germ cells with a sufficient number of mitochondria can be induced from iPS cells.

#### **3. The Induction of Germ Cells from iPS Cells**

Human iPS cells were initially generated from somatic cells by the ectopic expression of four transcription factor genes (OCT4, SOX2, KLF4, and c-MYC) in 2007 [29]. The current iPS cell generation methods vary in the choice of somatic origin, the set of reprogramming factors, and the transduction methodology [30]. The new pluripotent stem cells have become the starting material for germ cell induction, in which ES cells had been used (Table 1). Clinical applications of iPS-derived germ cells require scientific scrutiny in terms of meiosis, epigenetic programming, and the organization of the nucleus and mitochondria. Based on lessons learned from previous research on human ES cells [3–8] (Table 1), non-human primate ES cells [31], and mouse pluripotent stem cells [1,2,32–35], the current primary differentiation strategy involves differentiating human iPS cells into PGCs, and subsequently directing the PGCs to undergo meiosis, with some variations (Table 1). The PGC formation has been verified by the expression of marker genes or immunostaining for marker proteins including VASA (DDX4), cKIT, and SSEA1 (Figure 2). Confirming entrance into meiosis involves assessing the haploidy of differentiated cells as well as detecting meiosis-associated markers, such as acrosin, transition protein 1 (TP1), and protamine 1 (Prot1).


**Figure 2.** Differentiation pathway from human iPS cells to germ cells. Human iPS cells are differentiated into primordial germ cells (PGCs), and further differentiated into meiotic cells. Indicated information regarding confirmed markers is derived from research reports regarding germ cell induction using human iPS cells. PGCs: primordial germ cells, SSCs: spermatogonial stem cells.

*3.1. Induction of the Male Germline* 

The differentiation of human male iPS cells has so far produced PGCs [8–10], gonocytes [11], SSCs [12], spermatocytes [12,13], and haploid, round spermatid-like cells [10,12,13] (Figure 2, Table 1). As early as 2009, Park *et al.* [8] reported PGC induction from human iPS and ES cells. They used a triple biomarker (cKIT, SSEA1, VASA) assay to identify and isolate the PGCs, and demonstrated that culturing such human pluripotent stem cells on human fetal gonadal stromal cells, which were derived from a 10-week-old human fetus, significantly improved the efficiency of PGC formation. Moreover, the efficiency was comparable among various ES cell and iPS cell lines. Utilizing bisulfite sequencing, they showed that the PGCs initiate imprint erasure from differentially methylated imprinted regions (H19, PEG1, and SNRPN DMRs) by day seven of differentiation. However, PGCs derived from iPS cells did not initiate imprint erasure as efficiently, suggesting that further investigation is needed on the epigenetic status during germ cell induction from iPS cells.

In 2011, Panula *et al.* compared the potential of human iPS cells, derived from adult and fetal somatic cells to form primordial and meiotic germ cells [10]. As a consequence, approximately 5% of human iPS cells were found to have differentiated into PGCs with induction by bone morphogenetic proteins (BMPs). In addition, by overexpressing intrinsic regulator genes, including *DAZ*, *DAZL*, and *BOULE*, iPS cells formed meiotic cells with extensive synaptonemal complexes and post-meiotic haploid spermatid-like cells. These results show that human iPS cells generated from adult somatic cells can form germline cells other than PGCs. More recently, similar results using the overexpression of *VASA* and/or *DAZL* was reported, demonstrating that both human ES cells and iPS cells differentiated into PGCs, and the maturation and progression of these cells through meiosis was enhanced [9]. Again, post-meiotic male haploid cells were induced in 14 days following the overexpression of the two regulators. Moreover, the methylation pattern of the H19 locus, similar to that of normal germ cells, was observed following the expression of *VASA* alone. Therefore, such RNA-binding proteins appear to promote the meiotic progression of human iPS cell-derived germ cells *in vitro*.

In contrast to these studies, Eguizabal *et al.* demonstrated that without the overexpression of germline related factors, postmeiotic haploid cells were consistently obtained from human iPS cells of different origins (keratinocytes and cord blood), generated with a different number of transcription factors [13]. Their two-step differentiation protocol begins with iPS cell culture for three weeks with human ES cell media in the absence of bFGF. Subsequently, retinoic acid (RA) is added to the medium, and the culture continues for three more weeks. Then, the cells are sorted and reseeded onto culture plates in the presence of forskolin (FRSK), human recombinant leukemia inhibiting factor (rLIF), bFGF, and the CYP26 inhibitor, R115866, for at least two weeks. Consequently, the post-meiotic spermatid-like cells with acrosin-staining were identified. Moreover, Easley *et al.* also reported a similar direct differentiation approach without the overexpression of genes [12]. They adopted standardized mouse SSC culture conditions [36] and demonstrated that human ES cells and iPS cells differentiated directly into advanced male germ cell lineages, without genetic manipulation. They observed spermatogenesis *in vivo* by differentiating these pluripotent stem cells into UTF1-, PLZF-, and CDH1-positive spermatogonia-like cells; HIWI- and HILI positive spermatocyte-like cells; and haploid, round spermatid-like cells expressing acrosin, TP1, and Prot1. Such spermatids had uniparental genomic imprints similar to those of human sperm on two loci: H19 and IGF2. These results demonstrate that male iPS cells have the ability to differentiate directly into haploid, round spermatids *in vitro*.

Therefore, male germ cell induction from iPS cells has rapidly advanced since 2009. Although transplantation of autologous SSCs to restore spermatogenesis has already succeeded in infertile monkeys [37], clinical use of SSCs induced from iPS cells requires considerable caution. Notably, Amariglio *et al.* warned that transplantation of stem cells, not differentiated cells, in a patient could cause an adverse event [38]. They reported that a boy with ataxia telangiectasia treated with the intracerebellar and intrathecal injection of human fetal neural stem cells was diagnosed with a multifocal brain tumor four years after the first injection. One might consider using iPS cell-derived spermatids in the clinical setting. However, although oocytes have been fertilized with elongated spermatids [39,40], they were insufficiently fertilized with premature, round spermatids, resulting in poor embryonic development [41–43]. Successful fertilization of oocytes with more matured male germ cells *in vitro* needs to be examined in preclinical research.

Based on the recent mouse work by Hayashi *et al.* [1], primordial germ cell-like cells (PGCLCs) were generated from ES cells and iPS cells through epiblast-like cells (EpiLCs), a cellular state highly similar to pregastrulating epiblasts but distinct from epiblast stem cells (EpiSCs). To examine whether such PGCLCs undergo proper spermatogenesis, PGCLCs were transplanted into the seminiferous tubules neonatal mice lacking endogenous germ cells. As a result, fertile spermatozoa were produced in the thick tubules. The global transcription profiles, epigenetic reprogramming, such as imprinted genes (*Igf2r*, *Snrpn*, *H19*, and *Kcnq1ot1*), and cellular dynamics during PGCLC induction from EpiLCs resembled those associated with PGC specification from the epiblasts. Remarkably, they identified Integrin-b3 and SSEA1 as markers that could be used to isolate PGCLCs from differentiated cells. More recently, Ramathal *et al.* demonstrated that human iPS cells transplanted directly into mouse seminiferous tubules differentiated extensively to form germ cell-like cells with morphology indistinguishable from that of fetal germ cells, and these cells expressed PGC-specific proteins including VASA, DAZL, and STELLA [11].

These findings revealed the significance of differentiation pathway from iPS cells to germ cells and elaborated the need for culture conditions that mimic the stem cell niche in the testis to efficiently and effectively direct human iPS cells to form more advanced germ cells *in vitro*.

#### *3.2. The Induction of Female Germline*

In contrast to male germline induction, the differentiation of iPS cells or ES cells into female germ cells has been insufficiently studied (Table 1, Figure 2). Eguizabal *et al.* consistently observed between 1.0%–2.0% haploid cells per human female iPS cell line (derived from keratinocytes or cord blood) in their two-step differentiation protocol [13]. Their female iPS cells were differentiated into haploid cells following the detection of the SCP3 and H2AX proteins (indicators of meiotic competence). However, they observed that most of the iPS cell lines, including female cells, increased their methylation status of H19 (the maternally expressed, paternally imprinted gene), displaying a clear tendency toward paternal imprinting. Therefore, it appears that the germ cells induced from female iPS cells are certainly haploid, but are incomplete as mature female germ cells because oocytes only extrude the last polar body after fertilization. Panula *et al.* also reported a similar result regarding the differentiation of female iPS and ES cells into meiotic germ cells by the overexpression of the intrinsic regulators [10]. Moreover, Bucay *et al.* showed that germ cells differentiated from human ES cells *in vitro* express both male and female genetic programs regardless of their karyotype [7].

With regard to mouse systems, there have been attempts to induce female germ cells from ES cells since 2003 [44–48]. Although a follicle-like structure with oocyte-like cells was spontaneously observed, entrance into meiosis was not confirmed in those reports. Nicholas *et al.* clearly noted that mouse ES cell-derived oocyte maturation ultimately fails *in vitro* [48]. They transplanted ES cell-derived oocyte-like cells into an ovarian niche to direct their functional maturation and showed that the physiological niche of the ovary is required for their differentiation. Notably, Hayashi *et al.* showed that mouse female ES cells and iPS cells were differentiated into fertile oocytes via EpiLCs and PGCLCs [2], using a combined *in vitro* and *in vivo* system which led to the successful induction of fertile spermatozoa in 2011 [1]. When the PGCLCs were aggregated with female gonadal somatic cells as reconstituted ovaries, they underwent X-reactivation, imprint erasure, and cyst formation, and exhibited meiotic potential. After PGCLCs in the reconstituted ovaries were transplanted under the mouse ovarian bursa, such cells matured into germinal vesicle-stage oocytes, and contributed to fertile offspring after *in vitro* maturation and fertilization. Therefore, the differentiation of female iPS cells into germ cells largely depends on the use of an *in vivo* system.

The difficulty in female germ cell induction from pluripotent stem cells is more likely to reflect *in vivo* oogenesis. Human gametogenesis initiates around the 23–26th day post-conception [49]. The precursors of gametes, the PGCs, appear in the dorsal wall of the yolk sac near the developing allantois. The PGCs proliferate and migrate through the dorsal mesentery into the gonadal ridges. The PGCs are found in the gonads by the fourth week post-conception. Thereafter, female and male PGCs differentiate into oogonia (subsequently, oocytes) or gonocytes (subsequently, spermatozoa), respectively. The male germ cells undergo mitotic arrest until birth, whereas the female germ cells further enter meiotic arrest (Figure 2). Following birth, such germ cells are reactivated and resume meiosis, resulting in the beginning of the production of mature oocytes and spermatozoa after puberty. Therefore, human gametogenesis proceeds on a long-term basis with gender differences in meiotic progression.

In males, SSCs are maintained, and contribute to spermatogenesis by self-renewal *in vivo* for a long time. In addition, human SSCs can be maintained *in vitro* for a long term. Sadri-Ardekani *et al.* demonstrated that the human SSC numbers increased 53-fold within 19 days in testicular cell culture and increased 18,450-fold within 64 days in a germline stem cell subculture [50]. Conversely, it has generally been considered that most female germ cells enter meiosis I until birth, do not proliferate after birth, and that the number of the germ cells gradually declines until menopause (at approximately 40 years) [51]. The significant differences in the proceedings between spermatogenesis and oogenesis appear to impact the differentiation of human iPS cells into germ cells in the laboratory. However, there have been several unique reports regarding mammalian oogenesis. Some groups have reported the isolation of oogonial stem cell-like cells in mice and humans [52–55]. However, there are counterarguments about the existence of oogonial stem cells [56–58]. If the mitotically active oogonial cells can be isolated in a reproducible manner, the findings are expected to contribute to enhancing female germ cell induction as well as providing a mitochondrial resource for ooplasmic transfer.

#### **4. Future Directions**

In order to improve the induction efficiency and functional completeness of germ cell induction from human iPS cells, deeper insight into iPS cell generation and gametogenesis *in vivo* is vital. In addition, creating human embryos is likely to require the assessment of the developmental potential of induced germ cells. The conditions to permit the creation of human embryos for these functional assays should be discussed, because such experiments are frequently associated with ethical concerns or issues [15].

#### *4.1. Genetic and Epigenetic Stability of Human iPS Cells*

As mentioned in the Introduction, the ART using induced germ cells appears to be against the Weismann barrier. Induced germ cells are likely to be subject to genetic and/or epigenetic instabilities during iPS cell generation and germ cell induction. The genetic stability of iPS cells significantly impacts their research use, in addition to their safe medical use. Some cytogenetic analyses have suggested that human iPS cells and ES cells are likely to acquire trisomies in chromosome 12, and 17, indicating an underlying mechanism of growth advantage associated with culture adaptation [59–61]. Moreover, the tendency for large-scale chromosomal aberrations appears to have no dependence on the cell origin or iPS generation methods, although some of the chromosomal aberrations observed in PS cells were derived from the original somatic cells [59,60,62]. In addition, human iPS cell cultures are likely to undergo chromosomal changes at both early and late passages. A close examination of the genetic changes during culture indicated that the observed peak in occurrence of chromosomal aberrations is at around passage eight in iPS cells, while that in ES cells is at around passage 36 [59]. Moreover, smaller copy number variations (CNVs) in human iPS cell culture are present across chromosome 12, 17, and 20 [63]. Compared with human ES cells, iPS cells showed increased CNVs, and had more CNVs at low passages (18%) than at late passages (9%) [62,64]. Therefore, human iPS cells seem to be subject to genetic changes at earlier culture stage, mostly resulting from somatic cell reprogramming.

The genetic instabilities might occur not only in nuclear DNA, but also in mitochondrial DNA (mtDNA). The copy number of mtDNA, which encodes proteins required to produce ATP for motility of spermatozoon ranges from 2.8 to 226 copies of mtDNA [65]. In contrast, in oocytes, the mtDNA copy number ranges from 20,000 to 598,000 [66–69], significantly impacting the outcome of fertility in ART. Since proper ATP production by mitochondria is essential for accurate meiosis in oogenesis as well as normal embryonic development [66–69], the mtDNA integrity of human iPS cells needs to be addressed. Relative to the founder fibroblasts, a higher rate of heteroplasmic variation was observed in human iPS cells [70]. Although this phenomenon may imply an increased mutation load in the iPS cells, such iPS cell lines showed no significant metabolic differences. Van Haute *et al.* tested 16 human ES cell lines and showed that they carry a plethora of diverse mtDNA deletions [71]. The mtDNA mutations did not seem to correlate with the time in culture, and were detected in the early passage cells. Such deletions did not appear to impact the differentiation potential, and were still present in terminally differentiated cells. Conversely, Wahlestedt *et al.* reported a unique result using a mutator mouse model with an error-prone mtDNA polymerase [72]. They investigated the impact of an established mtDNA mutational load regarding the differentiation properties of mouse iPS cells. As a consequence, the mutator iPS cells displayed delayed proliferation kinetics and harbored extensive differentiation defects, although somatic cells with a heavy mtDNA mutation burden were amenable to reprogramming into iPS cells. These findings suggest the need for careful analyses of the nuclear DNA and mtDNA in human iPS cells prior to germ cell induction.

In addition, epigenetic aberrations in human iPS cells have been pointed out, indicating defects in DNA methylation, including regions subject to imprinting [73]. Interestingly, high-resolution DNA methylation profiles suggested that some iPS cell lines possess somatic memory [74,75]. Although iPS cell lines with such memory might readily differentiate into germ cells, careful assessment of the epigenetic status of human iPS cells is required to avoid a low efficiency differentiation or aberrant epigenetics in the resulting germ cells.

#### *4.2. The Pluripotency State of Human iPS Cells*

Human ES and iPS cells are more similar to mouse EpiSCs that were derived from epiblasts in postimplantation embryos than mouse naive, ground state ES cells [76,77]. The features of the ground state pluripotency include driving *Oct4* (also known as *Pou5f1*) transcription by its distal enhancer, globally reduced DNA methylation, prominent deposition of the repressive histone modification H3K27me3, and bivalent domain acquisition on lineage regulatory genes [78]. Moreover, human female ES and iPS cells frequently show a pronounced tendency for X chromosome inactivation. These lines of evidence suggest that human iPS cells represent a primed state of pluripotency that is distinct from the naive pluripotent ground state of mouse ES and iPS cells. Recently, some new methods to establish human iPS cells have been proposed [79–81]. These methods, which are based on 2i/LIF conditions (exogenous stimulation with leukemia inhibitory factor and small molecule inhibition of ERK1/ERK2 and GSK3ȕ signaling) with additional components, demonstrated the establishment of human iPS cells in the naive ground state [79,81], or in the preimplantation epiblast state [80]. The use of human iPS cells generated by such methods are likely to facilitate the subsequent appropriate differentiation pathway to germ cells, as demonstrated by the two mouse experiments in which mouse pluripotent stem cells were differentiated into germ cells via EpiLCs [1,2].

#### *4.3. Spatio-Temporal Factors in Gametogenesis*

Currently, germ cell induction from human iPS cells is advancing primarily in the male germline. A better understanding of gametogenesis would facilitate the induction of female germ cells, as well as the terminal differentiation into spermatozoa. Following puberty, spermatogenesis occurs at the seminiferous tubules in the testis in which Sertoli cells co-exist with Leydig cells. In inducing male germ cells, co-culture with Sertoli cells that foster and differentiate spermatocytes *in vivo* has already been introduced to induce spermatogenesis *in vitro*. Park *et al.* improved PGC generation using a co-culture system with human fetal gonadal cells [8]. Moreover, Bucay *et al.* reported that PGC generation from human ES cells was accompanied by the development of Sertoli-like support cells [7]. Moreover, another article reported that testosterone, which the Leydig cells of the testes produce, was added to the culture medium in order to promote differentiation of mouse iPS cells into male germ cells *in vitro* [82]. More elaborate culture systems including Sertoli cells and Leydig cells may be effective to induce terminally differentiated male germ cells. However, fetal and adult populations of Leydig cells are distinct cells in terms of their physiology and function [83]. A recent report suggested that Sertoli cells support adult Leydig cell development in the prepubertal testis [84]. Regarding female germ cell development, oocytes are surrounded by a single layer of flattened ovarian follicular epithelial cells at meiotic arrest. When stimulated at puberty, the oocyte enlarges, and the follicular cells continue to proliferate to form many layers surrounding the oocyte. These cells eventually become known as granulosa cells that secrete progesterone after ovulation. Qing *et al.* have used co-culture with ovarian granulosa cells in the induction of oocyte-like cells expressing oocyte-specific genes including Figalpha, GDF-9, and ZP1-3 from mouse ES cells [47]. Interestingly, when they were co-cultured with Chinese hamster ovary (CHO) cells or cultured in CHO cell-conditioned medium, these cells did not express all of these oocyte-specific markers during the germ cell induction. Moreover, Nicholas *et al.* differentiated mouse Oct4-GFP ES cells *in vitro*, isolated GFP positive germ cells by FACS and co-aggregated the cells with dissociated mouse newborn ovarian tissue [48]. Subsequently, they transplanted the co-aggregates under the kidney capsule of recipient mice. They observed ES cell-derived Oct4-GFP positive oocytes in the graft despite the efficiency being low. Furthermore, in a recent work [2], Hayashi *et al.* differentiated PGCLCs, which were induced from mouse ES and iPS cells into fertile oocytes, by using *in vitro* aggregation with female gonadal somatic cells and transplantation of germ cells under the mouse ovarian bursa. The spatio-temporal factors associated with human gametogenesis *in vivo* should be further considered to develop more elaborate culture or differentiation systems in order to increase the possibility of inducing more mature germ cells from human iPS cells.

#### *4.4. Assessing the Developmental Potential of Induced Germ Cells*

In order to confirm whether induced human germ cells possess the correct biological functions, creating embryos and culturing them for a short term is indispensable prior to considering the use for clinical applications. In doing so, a subsequent biological analysis would necessitate the establishment of ES cells from the embryos. Nonetheless, these experiments are likely to raise ethical concerns owing to the fact that such embryos are created and destroyed for research purposes, not for reproduction. In some countries, creating a human embryo and monitoring the development of human embryos until the 14th day post-conception or until the beginning of the formation of the primitive streak may be permitted with approval of an institutional review board (IRB) and/or national authorities [15]. However, such human embryo experiments require sufficient data to support their use based on animal experiments to confirm scientific or medical rationality. Since non-human primate (NHP) experiments are more scientifically comparable with the human conditions than experiments in lower animals such as rodents, the data obtained from NHP experiments are likely to be required by IRB or other bodies with regard to granting permission for human embryo research.

#### **5. Conclusions**

As discussed above, human germ cell induction has advanced primarily in the male germline, progressively reaching to a final differentiation stage. Meticulously selecting human iPS cell lines with higher pluripotency and genetic integrity is expected to improve the efficiency of the formation of PGCs and entrance into meiosis. Moreover, placing the differentiated cells in culture systems similar to the niche in human gonadal tissues will likely produce not only spermatozoa, but also female germ cells that are more similar to oocytes. Further considerations of the intrinsic regulators that could be overexpressed are also likely to advance meiotic progression, complete meiosis, and functionally mature these germ cells.

Rapid advances in stem cell research will likely enable human iPS cells to differentiate into elongated spermatids or *bona fide* spermatozoa within the next decade or less. Recently, perplexing ethical and social concerns associated with the careless use of induced germ cells have been raised [15]. The use of ART with induced germ cells might facilitate posthumous conception, the birth of many siblings in a region without their knowing their genetic relationships, and facilitating the birth of a "savior sibling" to provide HLA-matched transplantation therapy for a relative. The uncontrolled or unethical use of induced germ cells would make the current problems associated with ART more complicated. As human germ cell induction from human iPS cells proceeds, appropriate deployment of this stem cell technology in ART will become an urgent matter that will need to be addressed by both researchers and the general public, including prospective parents.

#### **Acknowledgments**

The author thanks Motoko Araki for supporting figure drawing. This work was supported by the Japan Society for the Promotion of Science (JSPS) KAKENHI Grant Number 26460586 (TI).

#### **Appendix**

#### **Assisted Reproductive Technology (ART)**

ART involves several types of medical procedures to achieve pregnancy. Types of ART include IUI, oocyte retrieval, IVF, and ICSI.

#### **Intrauterine Insemination (IUI)**

At the early stage of ART, IUI is performed by placing spermatozoa inside a woman's uterus in order to facilitate fertilization.

#### **In Vitro Fertilization (IVF)**

IVF begins with the induction of ovulation via hormonal stimulation, followed by oocyte retrieval. Subsequently, the retrieved oocytes are fertilized with spermatozoa in a petri dish. The resulting embryos are cultured for three to five days following fertilization, and one or more viable embryo is transferred to the uterus.

#### **Intracytoplasmic Sperm Injection (ICSI)**

In cases of male infertility, one spermatozoon is generally injected into an oocyte to facilitate fertilization under a microscope. The embryos are cultured and transferred as in IVF.

#### **Ooplasmic Transfer [23,27,28]**

In cases of female infertility, ooplasm, including mitochondria from fresh, mature or immature, or cryopreserved-thawed donor oocytes are directly injected into recipient oocytes via a modified ICSI technique to enhance the viability of the oocytes. Currently, there is a moratorium on this procedure in the U.S. and Canada due to the potential health risks to the progeny.

#### **Conflicts of Interest**

The author declares no conflict of interest.

#### **References**


Chapter9: Genetic Disorders

### **Comparing ESC and iPSC—Based Models for Human Genetic Disorders**

#### **Tomer Halevy and Achia Urbach**

**Abstract:** Traditionally, human disorders were studied using animal models or somatic cells taken from patients. Such studies enabled the analysis of the molecular mechanisms of numerous disorders, and led to the discovery of new treatments. Yet, these systems are limited or even irrelevant in modeling multiple genetic diseases. The isolation of human embryonic stem cells (ESCs) from diseased blastocysts, the derivation of induced pluripotent stem cells (iPSCs) from patients' somatic cells, and the new technologies for genome editing of pluripotent stem cells have opened a new window of opportunities in the field of disease modeling, and enabled studying diseases that couldn't be modeled in the past. Importantly, despite the high similarity between ESCs and iPSCs, there are several fundamental differences between these cells, which have important implications regarding disease modeling. In this review we compare ESC-based models to iPSC-based models, and highlight the advantages and disadvantages of each system. We further suggest a roadmap for how to choose the optimal strategy to model each specific disorder.

Reprinted from *J. Clin. Med*. Cite as: Halevy, T.; Urbach, A. Comparing ESC and iPSC—Based Models for Human Genetic Disorders. *J. Clin. Med.* **2014**, *3*, 1146–1162.

#### **1. Introduction**

Pluripotent stem cells have an unlimited self-renewal capacity and can differentiate into virtually any adult cell type [1] and even some extra-embryonic tissues [2,3]. These features make human pluripotent stem cells (hPSCs) a useful tool for disease modeling, which overcomes limitations observed in animal and adult human cellular models. While the use of animal models proved to be extremely valuable and successful in many cases [4], there are numerous diseases, such as Lesch-Nyhan syndrome [5], Turner syndrome [6] and Fragile X syndrome [7], that cannot be studied using animal models due to species-specific differences. The use of mature cells from patients can solve the species-specificity issue but this strategy is limited by the fact that it enables studying only a few types of cells at a specific developmental stage, and in many cases requires also transformation of the cells to enable their proliferation in culture. By contrast, due to their unique properties, hPSCs enable exploration of different types of cells, to study the effect of a specific mutation on differentiation or development and can proliferate *in vitro* without additional transformation. Indeed, since the generation of the first human embryonic stem cells (ESCs) based model (a model for Lesch-Nyhan syndrome by targeting of the *HPRT* gene in human ESCs) [5] dozens of disease models were generated by reprogramming of somatic cells from patients [1], by derivation of mutant ESCs from affected embryos diagnosed by *pre*-implantation genetic diagnosis (PGD) or by genetic manipulation of normal ESCs [8] (see Figure 1). While some models were used as a "proof of concept" to demonstrate that hPSCs can be derived from a wide range of disorders [9–11] or to show the feasibility of the mutant pluripotent cells to be used as a disease model [12], other models were further used to obtain novel mechanistic or physiological insights regarding the disorders. One example is a model for Amyotrophic Lateral Sclerosis (ALS) by Kiskinis *et al.* [13].

**Figure 1.** Human pluripotent stem cell-based models for genetic disorders can be generated by different techniques. Mutated human pluripotent stem cells can be derived by genetic manipulation of normal pluripotent stem cells, from affected embryos (identified by PGD), or from adult patients (by reprogramming of somatic cells).

The general differences between ESCs and induced pluripotent stem cells (iPSCs) and the utilization of hPSCs for disease modeling has been discussed extensively in the literature [1,14–16]. In this review we will focus on the differences between ESC-based models and iPSC-based models, and discuss the effect of genome editing technologies on the field of disease modeling.

#### **2. ESCs** *vs.* **iPSCs in Disease Modeling**

Theoretically, a given disorder can be equally modeled by iPSCs and by ESCs, as both are pluripotent stem cells. However, several reasons have made iPSCs derived from patients the system of choice:

(1) The use of normal human ESCs to model a genetic disorder requires genetic manipulation to induce the specific mutation that one would like to study. The way to obtain a mutation that will be identical to the natural occurring mutation, seen in patients, is by genome editing. However, the efficiency of genome editing in human ESCs, before the establishment of gene targeting technologies as discussed below, was extremely low (especially in cases where a homozygous mutation was required) [17] and derivation of iPSCs that already contain the specific mutation obviates the needs for this inefficient process.

(2) While the above mentioned limitation can be overcome by derivation of mutant ESCs from affected embryos identified by Preimplantation Genetic Diagnosis (PGD), this procedure is limited to a small number of diseases in which PGD is normally preformed, and can be done only in labs that are associated with *in vitro* fertilization (IVF) units.

(3) By contrast to iPSCs from affected individuals, in the case of ESCs based models, the correlation between the genotype and the phenotype is not obvious, and the penetrance of the mutation might be low as a results of specific "protective" genetic background [18].

(4) Lastly, in some countries the use of human ESCs is limited or banned due to ethical and religious concerns regarding the use of human embryos for research purposes as was discussed by others [19,20].

Nevertheless, possible drawbacks in modeling genetic disorders by iPSCs suggest that some disorders or specific aspects within a given disease might be better modeled in ESCs than iPSCs. The generation of a faithful iPSC-based model might be disrupted due to the following reasons (see Figure 2): (1) Incomplete reprogramming as a result of "Epigenetic memory" of the original somatic cells [21–23]; (2) Mutations accumulated during the reprogramming process [24] and deleterious effects (such as chromosomal instability, [25]) of the reprogramming process on the genome integrity of iPSCs; (3) Genetic aberrations that significantly decrease the reprogramming efficiency [26]; (4) The absence of appropriate sources of somatic cells such as in the cases of genetic aberration and aneuploidies that lead to very early embryonic lethality [6].

To demonstrate the commonalities and differences between ESC- and iPSC-based models, we compared models for X-linked, autosomal recessive and autosomal dominant disorders in which disease-related phenotypes were observed in both models (Table 1). As expected, in some cases the ESC-based models and the iPSC-based models were similar (spinal muscular atrophy [27,28], Shwachman-Dimond syndrome [29], long QT syndrome [30], and some aspects of myotonic dystrophy [31,32]). However, in other cases the iPSCs were limited in their capacity to model the disorder or specific aspects within the disorders. To demonstrate some of these cases, and to discuss the principles behind them, we will focus on the following disorders: Turner syndrome, Fanconi Anemia, fragile X syndrome and Huntington's disease.

**Figure 2.** Limitations in the generation of iPSC-based disease models. The "X-axis" in this scheme depicts the specific stages during the formation of iPSC-based models that might be affected by the different factors that discussed in the main text.

#### **3. Turner Syndrome**

X chromosome monosomy (XO) is one of the most common chromosomal abnormalities, as 3% of all pregnancies start with XO embryos [33]. Yet, approximately 99% of the XO embryos undergo miscarriage during the first trimester [33,34]. The 1% that survive to term are born with Turner syndrome which is characterized by several phenotypes; the most common among them are growth failure, gonadal dysgensis and webbed neck [34]. While Turner syndrome derived iPSCs can be used in order to study the phenotypes of the patient (pending the availability of the required differentiation protocols), they might be problematic in modeling the early lethality of XO embryos, as they represent the exceptional 1% of the cases that survived to term.

In agreement with this notion, gene expression analysis of XO ESCs (derived by screening for ESCs with normal karyotype that lost one of their X chromosomes) revealed a significant effect of X chromosome monosomy on the expression of placental genes and suggests that the reason for the early lethality is abnormal placental development [6]. By contrast, there was almost no effect of X chromosome monosomy on placental gene expression in iPSCs derived from Turner syndrome patients, and even from amniotes of a 20 weeks old embryo [35]. The results suggest that Turner syndrome iPSCs represent the rare cases in which the embryo survived despite the XO karyotype.





#### **4. Fanconi Anemia**

Fanconi Anemia (FA) is an autosomal recessive disorder caused by a mutation in any of the 16 *FANC* genes and characterized by congenital abnormalities, cancer predisposition and progressive bone marrow failure [44]. Initial attempts to reprogram somatic cells from FA patients into iPSCs failed unless using fibroblasts that were first genetically corrected [45]. These results suggested that the FA pathway is essential for the reprogramming pathway (probably due to defective DNA repair and genomic instability of FA cells) and therefore that FA can't be easily modeled by iPSCs. However, further attempts to reprogram "uncorrected" somatic cells from FA patients under hypoxic conditions [26], and even under normoxic conditions [25], showed that iPSCs can be derived from FA somatic cells, albeit in a very low efficiency and revealed that "…somatic cells harboring mutations that render the FA pathway defective are resistant but not refractory to reprogramming" [26]. Nevertheless, significant chromosomal aberration in uncorrected FA-iPSCs [25], but not in FA-iPSCs derived from "corrected" somatic cells [26] or in human ESCs with stable knockdown of FANCC [25] suggests that the FA pathway is required to prevent DNA damage and chromosomal instabilities associated with the reprogramming process. The severe aneuploidy in the uncorrected FA-iPSCs but not in the ESC-based model for Fanconi anemia suggests that ESCs and not iPSCs should be used to study FA. Surprisingly though, it has been recently shown [41] that FA-iPSCs with a normal karyotype can be derived from FA somatic cells upon episomal reprogramming. Moreover, the FA-iPSCs were very similar to FA-ESCs that were generated by gene targeting of the *FANCA* gene using the TALEN mediated gene targeting [38]. The FA-iPSCs and the FA-ESCs were extensively studied and compared to isogenic control cells (the original ESCs and target corrected FA-iPSCs) and proved to be a very useful model for different aspects of Fanconi anemia. While the reasons for the differences in the chromosomal stability between the viruses mediated reprogramming and the integration-free episomal mediated reprogramming are still not clear, these results indicate that in some cases, the reprogramming method itself might have a dramatic effect on the quality of the iPSCs and thus, should be taken under consideration when choosing to generate a disease model by reprogramming of somatic cells from patients.

#### **5. Fragile X Syndrome**

Fragile X syndrome (FXS) is a trinucleotide repeat disorder and is the leading cause of inherited intellectual disability in males, affecting approximately one in every four thousand boys and one in eight thousand girls worldwide [46–49]. The mutation leading to the syndrome is a trinucleotide CGG expansion at the 5ƍ untranslated region of the fragile X mental retardation 1 (*FMR1*) gene, which is accompanied by epigenetic changes, resulting in the silencing of the gene [49,50]. The product of the *FMR1* gene is the fragile X mental retardation protein (FMRP) which is most abundant in the brain and testis and plays a major role in synaptic plasticity [51].

In 2007 human ESCs from FXS affected embryos (FXS-ESCs) were derived for the first time through PGD and enabled the study of the development of the disease [7]*.* Interestingly, although carrying the full mutation, FXS-ESCs showed both *FMR1* mRNA expression and the presence of FMRP. This finding showed that the transcriptional silencing of *FMR1* is a developmentally regulated process. Moreover, the study indicated that *FMR1* is silenced in FXS embryos only during development and that the inactivation is initiated by chromatin modifications prior to DNA methylation [7]. Other studies on FXS-ESCs supported the finding that *FMR1* is expressed in full mutation embryos and is silenced only during differentiation and further demonstrated that *FMR1* plays an important role in early stages of neurogenesis and synaptic function [52,53]. Therefore, FX-ESCs are invaluable to study many aspects of FXS, first and foremost the epigenetic silencing mechanism. However, there are also limitations in the FX-ESCs model: FXS is represented by profound variability in patients, ranging from the varying length of the repeats, through the methylation levels, and to the neurological phenotype itself. The degree of intellectual impairment also varies between different individuals, as only about 30% of full mutation carriers display autistic behavior [54,55]. Additionally, some carriers of the full mutation allele do not display any of the syndrome's phenotypes [56,57]. As this variability is not inherited from the parents and is detected only after PGD analysis, the probability of acquiring numerous human embryonic stem cells displaying the entire spectrum of genetic and epigenetic differences is quite small and may take several years.

In contrast to the *FMR1* expression seen in FXS-ESCs, it seems that in FXS derived iPSCs (FXS-iPSCs), despite successful reprogramming of patients derived fibroblasts, the *FMR1* gene is resistant to the process and remains methylated and silent [36,58,59]. Thus, while the FXS model in human ESCs demonstrated the temporal silencing of *FMR1*, in FXS-iPSCs *FMR1* was already inactive in the undifferentiated state. This fundamental difference between FXS-ESCs and FXS-iPSCs controls the choice of model according to the question being asked. In order to better understand the different aspects of the initiating steps of the *FMR1* silencing such as CGG methylation and the epigenetic silencing, one should use the FXS-ESC model. On the other hand, if one wishes to model neural development, screen for new drugs or understand the CGG expansion mechanism it is preferential to use the FXS-iPSC model to understand the effects of lack of FMRP on developing neurons, as we do not fully understand at which time point during the differentiation process *FMR1* is silenced in ESCs *in vitro*. One example using FX-iPSCs to model Fragile X syndrome is a study aimed to evaluate the reactivation of *FMR1* in FXS-iPSCs and their neuronal derivatives through epigenetic modulation drugs. This study showed not only that reactivation is possible but also uncovered additional layers of epigenetic control on *FMR1* [60].

#### **6. Huntington's Disease**

Huntington's disease (HD) is an autosomal dominant neurological disorder caused by a trinucleotide repeat expansion and characterized by a late onset progressive neurodegeneration ending with death [61,62]. In HD, an expansion of a CAG repeat in the first exon of the Huntingtin (*HTT*) gene leads to a toxic gain of function activity of the mutant Huntingtin protein (mHTT), containing an increased number of polyglutamines at the *N* terminus [62]. These polyglutamine tails are then cleaved and accumulate as aggregates in the nuclei of neurons [63].

During the past few years, several groups have successfully created iPSC models for HD (HD-iPSCs) [11,43,64,65]. Some have further differentiated HD-iPSCs to neurons and showed increased caspase activity of neural precursors upon growth factor deprivation [64] or increased lysosomal activity in both HD-iPSCs and derived neurons [65]. The most comprehensive work done with the HD-iPSCs model system was performed by the HD-iPSC Consortium, in which several HD-iPSC lines were created and analyzed by a group of different labs [43]. In this work, HD-iPSCs were also differentiated into neural stem cells (NSCs) and neurons. HD-derived NSCs showed differential gene expression accompanied with changes at the protein level as well. Other changes observed were compromised energy metabolism, inability to fire action potential and increased cell death. Neurons also display increased death under different stress conditions most notably in lines containing longer repeats.

HD-iPSCs provide a useful model, however, it was never shown that they accumulate any insoluble aggregates, and thus cannot be used to study the formation and pathological contribution of this aggregates to the development of the disease. In order to study this aspect of the syndrome, normal ESCs were genetically engineered to express the polyglutamine repeats [42]. Neurons derived from these HD-ESCs matured over a period of several months and showed the polyglutamine aggregates. Similar to HD-iPSCs, HD-ESCs derived neurons exhibited progressive death under stress conditions. It was also shown using this model that reduction of mHTT by just 10% is sufficient to prevent toxicity and lowering the expression levels of *HTT* by up to 90% had no effect on neurons, opening the possibility to screen for new drugs to control the levels of mHTT. Thus, HD-ESCs may provide a stronger tool than HD-iPSCs in our understanding of the initiation and progression of the pathology of HD. However, work done on ESCs derived directly from an embryo with HD did not show the formation of polyglutamine aggregates [66]. Furthermore, HD embryos from PGD are not readily available, and due to the fact that HD is a late onset disease, we do not know the ultimate phenotype of these never developed embryos. In this case, more work should be done on both HD-ESCs and HD-iPSCs in an attempt to obtain more of the molecular phenotypes characteristic of the disease to create a better model system.

#### **7. Disease Modeling by Gene Targeting of hPSCs**

As mentioned above, hPSCs based models can be generated by the derivation of ESCs from affected embryos diagnosed by PGD or by genetic manipulation of normal hPSC cells. Down-regulation or over-expression of specific genes can be easily achieved by RNAi technologies (for down regulation) or by introduction of exogenous genes into the genome (for over-expression). While these methods proved to be very informative in some cases, they can't mimic the natural occurring mutation in the patients and therefore the relevance of the finding to the disease might be questionable in other cases. To overcome this problem, one has to induce a specific mutation that is identical to the mutation occurring in patients. However, until lately, genome editing in mammalian cells was an extremely inefficient process [17], and therefore it was challenging to generate homozygous mutations in human cells using the traditional methods for gene targeting. The development of new technologies for gene targeting, (reviewed in details in [17]) especially the TALEN technology and the Cas/CRISPER technology have dramatically increased the efficiency of gene targeting in mammalian cells and enabled to correct specific mutations or to obtained homozygous mutations in reasonable efficiency in human pluripotent stem cells.

These methods enable, for the first time, the comparison between isogenic cells that differ only in the specific mutation under investigation. This can be achieved by induction of a specific mutation in otherwise normal ESCs, by correction of a specific mutation in iPSCs or by combination of both methods [30,41] One possible drawback in these methods is the possible off-target effect that might result in additional unplanned genetic aberrations [17]. To overcome this possibility it is important to design the targeting sequence in a way that will decrease the off-target effects and to target different sequences of the same gene. Among these methods, the CRISPR technique will probably become the first choice for most labs due to the combination of accuracy, efficiency and accessibility.

#### **8. "Guidelines" for Choosing the Optimal Model System for a Given Disease**

The choice between modeling disorders with ESCs or iPSCs is dependent on several factors. We propose that the optimal model that probably overcomes most if not all the drawbacks mentioned above, is a model that combines both ESCs (that were genetically modified to carry a specific mutation) and iPSCs from patients (with an isogenic control of iPSCs from the same patient in which the mutation was corrected by genomic engineering). Such "combined methods" have been recently generated for long QT syndrome [30] and for Fanconi anemia [41]. However, as was discussed above, in some cases only one of the two methods is doable/informative. In Figure 3 we suggest general guidelines that should assist in choosing the right system for a given disorder. We generated this scheme based on the following assumptions:

**Figure 3.** Scheme depicting the steps in choosing the appropriate system for disease modeling. While in some cases there is only one possible option (either ESCs or iPSCs), in other cases both ESCs or iPSCs can be used and the decision between the two methods should be done after the consideration of the advantageous and disadvantageous of each one of the options (some of them are described in the scheme).

Multifactorial disorders in which there is a major contribution to factors other than genetic factors on the disease etiology can be modeled exclusively by iPSCs from patients that already manifested the disease and can't be modeled by ESCs. Similarly, iPSCs but not ESCs should be used to model multigenic disorders in which the genetic factor cannot be narrowed down into a single gene, (the discussion regarding the use of hPSCs to model these type of disorders is out of the scope of this review). Other than these two groups of disorders, "monogeneic" disorders can be modeled theoretically using both systems but under the following notions: (1) Genetic aberrations that lead to early lethality should not be modeled by iPSCs that derived from individuals that survived to term as they represent the exceptional cases (these rare cases however, can be used to study the effect of the genetic aberration on the phenotype of the exceptional embryos that survived to term and the genetic backgrounds that enable them to escape the early lethality); (2) Possible epigenetic memory in iPSCs has to be taken under consideration. While in general the epigenetic memory is considered to have a negative effect on iPSCs (as the pluripotent cells retain some of their previous identity of adult cells and therefore might not be equivalent to normal pluripotent cells), one can also utilize these phenomena in a positive manner. For example, in cases of hematopoietic disorders, iPSCs that were derived from blood cells might undergo hematopoietic differentiation in a greater efficiency than ESCs or iPSCs derived from other somatic cells [16]; (3) In cases in which no epigenetic effect is predicted, the best choice is to combine both model systems. When only one of the two systems will be used it is important to keep in mind the following limitations of each one of the methods: (a) The reprogramming process itself might results in accumulation of genetic aberrations that under some circumstances might affect the reliability of the model; (b) The penetrance of the mutation in the ESCs based model might not be completed (as a result of "protective" genetic background). By contrast, iPSCs are derived from patients that already manifested the phenotype and therefore one should not be concerned about "protective" genetic background; (c) In the case of PGD-based models the number of available samples (affected embryos) might be limited; (d) Gene targeting by the TALEN or CRISPER systems might lead to off-target effects.

#### **9. Conclusions**

Reprogramming of somatic cells from patients is a relatively easy procedure that doesn't involve the usage of human embryos, nor ethical issues, and results in the formation of iPSCs with the naturally occurring mutation. Therefore, since the first derivation of human iPSCs from normal donors [67–69] and from patients [11], this method was considered by many to be the optimal methodology for disease modeling by human pluripotent cells (due to scientific reasons as well as other reasons). Indeed, during the last several years numerous models for genetic disorders were generated by reprogramming of somatic cells from patients. Yet, in many cases the mutant cell lines were not further analyzed to study their relevance to the actual disorders. In this review we focused on iPSCs based models and ESCs based models that have been shown to have a phenotype related to the disease.

To demonstrate that iPSCs can't always replace ESCs in disease modeling, we focused on four models, each one emphasizes a specific aspect of the differences between ESC-based models and iPSC-based models. In addition to these specific examples, the reprogramming process itself might result in the generation of *de-novo* mutations [24] that might add "noise" to the system. On the other hand one general advantage of iPSC-based models compared to ESCs-based models is the fact that the patient chosen already manifested the phenotype associated with the mutation. This assures that there is no effect to the specific genetic background on the penetrance of the mutation. Based on the comparison between ESC-based models and iPSC-based models we suggested in Figure 3 a general guidelines to assist in choosing the appropriate model for a given disorder.

Lastly, the development of the "iPSCs technology" by Takahashi and Yamanaka some eight years ago [70], dramatically changed the entire field of pluripotent stem cells biology. While the most desirable application of this technology is probably for cell therapy, there is no doubt that currently the most common application of iPSCs is for disease modeling. In this review we highlighted some of the pros and cons of iPSCs compared to ESCs in regards to disease modeling and discussed the effect of advanced technologies for genome editing on the field. We believe that the field of disease modeling by hPSCs has reached a point wherein the challenge is not to derive pluripotent cells (ESCs or iPSCs) with a specific mutation but rather to better understand the pathophysiology of the disease and finding effective therapies.

#### **Acknowledgments**

The Authors would like to acknowledge Nissim Benvenisty for helpful discussions and critical comments on the manuscript.

#### **Author Contributions**

Tomer Halevy and Achia Urbach wrote the manuscript and designed the figures.

#### **Conflicts of Interest**

The authors declare no conflict of interest.

#### **References**



### **Design of a Tumorigenicity Test for Induced Pluripotent Stem Cell (iPSC)-Derived Cell Products**

#### **Shin Kawamata, Hoshimi Kanemura, Noriko Sakai, Masayo Takahashi and Masahiro J. Go**

**Abstract:** Human Pluripotent Stem Cell (PSC)-derived cell therapy holds enormous promise because of the cells' "unlimited" proliferative capacity and the potential to differentiate into any type of cell. However, these features of PSC-derived cell products are associated with concerns regarding the generation of iatrogenic teratomas or tumors from residual immature or non-terminally differentiated cells in the final cell product. This concern has become a major hurdle to the introduction of this therapy into the clinic. Tumorigenicity testing is therefore a key preclinical safety test in PSC-derived cell therapy. Tumorigenicity testing becomes particularly important when autologous human induced Pluripotent Stem Cell (iPSC)-derived cell products with no immuno-barrier are considered for transplantation. There has been, however, no internationally recognized guideline for tumorigenicity testing of PSC-derived cell products for cell therapy. In this review, we outline the points to be considered in the design and execution of tumorigenicity tests, referring to the tests and laboratory work that we have conducted for an iPSC-derived retinal pigment epithelium (RPE) cell product prior to its clinical use.

Reprinted from *J. Clin. Med*. Cite as: Kawamata, S.; Kanemura, H.; Sakai, N.; Takahashi, M.; Go, M.J. Design of a Tumorigenicity Test for Induced Pluripotent Stem Cell (iPSC)-Derived Cell Products. *J. Clin. Med.* **2015**, *4*, 159–171.

#### **1. Introduction**

Several notable clinical trials using human Pluripotent Stem Cell (PSC)-derived cell products have been conducted recently. In the first, Geron used embryonic stem cell (ESC)-derived oligodendrocyte progenitor cells (GRNOPC1) for treatment of acute spinal cord injury [1]. Advanced Cell Technology initiated a study in which ESC-derived retinal pigment epithelium (RPE) was used for treatment of Stargardt's disease and dry type Age-related Macular Degeneration (AMD) [2]. More recently, a clinical study for wet type AMD using induced Pluripotent Stem Cell (iPSC)-derived RPE was started at Riken CDB [3–5].

While clinical applications are moving forward, there are concerns that transplantation of differentiated PSC might lead to the formation of tumors in the recipient. Thus, examination of this possible outcome of transplantation is critically important. Cell transplantation or infusion therapy is distinctly different from drug administration. One must consider that transplanted or infused cells can survive for long periods in the host and may form tumors at the site of transplantation or at distal sites. The extent of tumor formation can be influenced by the microenvironment at the transplantation site or the ultimate homing site of the host. Furthermore, once a tumor has formed, it may influence the physical condition of the host through secreted factor(s) [6].

The aforementioned aspects of cell therapy must be addressed with animal transplantation studies prior to clinical use. Tumorigenicity tests that can assess the tumor-forming potential of transplanted cells are particularly important in the case of PSC-based cell therapies. As PSC have "unlimited" proliferation potential as undifferentiated stem cells, they can generate teratomas if they remain in the final product. The chance of generating a teratoma will increase if the procedure uses an autologous iPSC-derived cell product that presents no immunologic barrier. PSC might accumulate chromosomal abnormalities by selecting cells with unusual proliferative advantages over a long culture period. Lund *et al.* reported that some 13% of ESC and iPSC maintained in research labs worldwide demonstrated some type of genetic abnormality [7]. For that reason, the timely assessment of the genetic stability of PSC is of major interest for both research labs and clinical PSC banks. In addition, it is important to assess the potential for differentiation resistance due to incomplete reprogramming or a differentiation bias due to epigenetic memory when iPSC-based therapy is considered. In this context, it is necessary to assess the tumor-forming potential of non-terminally differentiated cells as well.

Information regarding genetic stability, gene expression, differentiation marker expression, cell growth rate and how cells were generated must be collected and evaluated prior to commencement of tumorigenicity testing. Next, it is necessary to have a clear idea about the scope and objective of related safety parameters: toxicology tests, Proof of Concept (POC) tests, biodistribution tests and tumorigenicity tests that can be conducted concurrently.

Toxicology tests can be designed depending on the properties of testing reagents and the purpose of the tests. The Organisation for Economic Cooperation and Development (OECD) Guideline for the Testing of Chemicals [8] is an internationally recognized test guideline for toxicology testing. They should be conducted in a blinded fashion to minimize the bias of measurement and observation by operators. Short-term and long-term end points are to be defined. Toxicology tests should be conducted by using clinically relevant methods of administration so that they can provide insights into a safe range of therapeutic cell doses. Acute (early) and late phase end points should be established in this test.

POC tests often employ a genetically modified animal that offers a model of the disease in question (e.g., Tg, KI, KO or KD mice) or injured animals to address the potential benefit or efficacy of the investigational therapy and to define the range of the effective dose used in clinical application by escalating the doses. The administration route and the method should be as close as possible to the intended clinical use. Positive and negative events should be clearly defined. In such a POC study, indices such as physiological recovery of lost function or overall survival of transplanted cells that could underlie intended therapeutic use are examined. Measurement of indices should be conducted in a blinded fashion to minimize bias during data acquisition. The size of the test group should be large enough to permit meaningful statistical analysis.

Biodistribution tests should be conducted to address tumorigenic proliferation of transplanted cells at the ectopic site. *Alu* sequence PCR is commonly used to detect human cells in host tissues or organs. While this PCR test detects human cells over a 0.1% frequency in host tissue by DNA ratio [9], greater sensitivity is needed to detect small metastatic colonies. In PET technology, proliferative cell mass is labelled by taking in a metabolic probe such as 18F FLT, providing a distribution of tumorigenic cell proliferation in the animal's body. However to trace the behavior of transplanted cells and their biodistribution over time requires labeling test cells by introducing marker genes by retrovirus or lentivirus that can emit a signal with a high S/N ratio. These approaches are currently under development.

#### **2. Guidelines for Tumorigenicity Tests**

Somatic cells with a normal chromosomal structure show limited proliferation potential. Tumorigenicity testing of mesenchymal stem cells may not reveal a serious problem [10]. However, in the case of PSC-derived cell products, the tumor-forming potential should be examined thoroughly because of the "unlimited" proliferation capacity of PSC and their genetic instability. However, there is no internationally recognized guideline for tumorigenicity testing of cells used for cell therapy. WHO TRS 878, "Recommendation for the evaluation of animal cell cultures as substrates for the manufacture of cell banks" [11,12] provides a guideline for animal cell substrates used for the production of biological medicinal products, but not for cells used for therapeutic transplantation into patients. Recently, FDA/CBER commented on the issues to be considered for cell-based products and associated challenges for preclinical animal study [13]. The report stated that when tumorigenicity testing of ESC-derived cellular products is undertaken, the tumorigenicity tests should be designed considering the nature of cell products to be transplanted and the anatomical location or microenvironment of the host animal. Tumorigenic test results from the administration of cells through nonclinical routes are not considered relevant as they would not assess the behavior of transplanted cells in the intended microenvironment to which the cells would be exposed. The study design should include groups of animals that have received undifferentiated ESCs, serial dilutions of undifferentiated ESCs combined with ESC-derived final products to infer the contamination of undifferentiated ESCs in the final product.

The aforementioned summarizes current discussions of tumorigenicity testing. However, we still need to answer a fundamental question: "How can we extrapolate animal tumorigenicity testing to humans?" The design of tumorigenicity tests should attempt to answer this question. For this, we must first estimate the risk that we will underestimate the incidence of tumor-forming events in humans by conducting an improper or non-informative animal study. So, how do we define such risk? For example, there is a risk that a study is unable to link unexpected tumor formation to genetic abnormalities of test cells presented before transplantation due to inadequate genetic information regarding test cells. In addition, there is a risk of obtaining "false" negative results by transplanting an insufficient dose, using an inadequate monitoring period, using an improper immunodeficient animal model that is insufficient to detect tumor, not transplanting into the right anatomical position, failure of transplantation itself or unexpected early death of transplanted cells in host tissue. We can address the risks by conducting quality control tests of test cells prior to transplantation and small scale pilot studies to determine the design of tumorigenicity tests. The following points should be considered in designing tumorigenicity tests.


#### **3. Specification of Test Cells**

Cells used in tumorigenic tests should be generated in a manner as close as possible to that intended for clinical use. In this context, it is preferable that cells used for all preclinical tests should be generated in a GMP-grade cell processing facility for clinical use. This approach would minimize bias originating from differences in cell production quality. Several types of data, including gene expression profiles obtained from gene chips or qRT-PCR to assess stem cell-like markers and differentiation markers, phenotypic analysis by flow cytometry, sterility tests, mycoplasma tests, exome sequencing, chromosomal stability tests with comparative genomic hybridization (CGH) array and karyotyping by multi-color banding (mBAND) or fluorescent *in situ* hybridization (FISH) would be valuable. For iPSC-derived cell products, EB formation assays would provide insights into differentiation potential. The results could be used to select "good" clones that demonstrate no differentiation bias or no differentiation resistance. These quality control tests and cell characterization tests are not a part of tumorigenicity testing *per se*. However, the information on starting material should be linked to the results of tumorigenicity testing to render the test results more informative.

In tumorigenicity testing of PSC-derived cell products, one can anticipate several tumor-forming events that include teratoma formation from residual "differentiation-resistant" PSC with normal karyotype, cancer-like progressive tumor formation from cells with abnormal karyotype or acquired genetic variation during culture and tumors with differentiation bias generated from imperfectly reprogrammed cells. To understand the nature of tumor-forming events, the link with results of these quality control tests is indispensable.

#### **4. Selection of an Animal Model**

In general, if one were to use "non-immunodeficient" healthy animals or "non-immunodeficient" disease model animals for tumorigenicity testing, one would have to administer a large amount of immunosuppressant for long-term monitoring. However, this approach will not always guarantee satisfactory engraftment of xeno-transplants. Primates can be used for tumorigenicity testing as models representative of humans, but this model is more useful for POC tests, not for tumorigenicity tests. Therefore, immunodeficient healthy rodents are widely used for tumorigenicity testing if human cells (final product) are to be used in the test. Large immunodeficient animals like the SCID pig [14] are also available. However, again, the SCID pig model would be useful to address transplantation efficiency of human cells, such as xeno-bone marrow transplantation of human hematopoietic stem cells as a part of a POC study in large animals. They are not cost-effective large scale statistical studies. To conduct tumorigenicity tests with a sufficient number of immunodeficient animals, a rodent model is a reasonable option for the preparation of test cells. Immunodeficient mice such as nude mice (BALB/cA, JCl-nu/nu), SCID mice (C.B-17/Icr-scid/scid), NOD-SCID mice (NOD/ShiJic-scid) and NOG mice (NOD/ShiJic-scid, IL-2RȖ KO) have been widely used for human cell transplantation studies. Prior to the design of tumorigenicity tests, one needs to evaluate the tumor-generating potential of these immunodeficient mouse strains by transplanting various dose of tumorigenic cell lines subcutaneously.

Another well-known transplantation site in rodents is beneath the testicular capsule space. This transplantation model is mainly used to test for satisfactory engraftment of test cells for POC tests, not for tumorigenicity tests. In our hands, it requires elaborate surgical skills and needs at least 104 iPSCs to generate tumors in NOG mice. In addition, tumor formation in the intraperitoneal space is hard to detect from the appearance of mice, thereby preventing statistical studies for tumor-forming events in a timely manner. In our case, the tumorigenic potential of immunodeficient mice was assessed by transplanting various doses of HeLa cells subcutaneously, following recommended procedure stated in WHO TRS 878 [11,12]. The mice were monitored over 12 months, and the TPD50 (minimum dose that can generate a tumor in 50% of transplanted mice) was calculated by the Trimmed Spearman-Karber method for each strain [9]. HeLa cells were used as a representative line of somatic tumorigenic cells with a genetic abnormality. For transplantation, a collagen-based gel lacking nutrients is sometime used to embed cells and to retain them at the designated transplantation site. Importantly, the gel *per se* does not support growth of the transplanted cells at the site. We have used Matrigel® (BD Biosciences, San Jose, CA, USA) to embed cells and to increase their tumor-forming potential [15]. We obtained the following values for the TPD50 for HeLa cells with Matrigel® via a subcutaneous route: Nude, 103.5 (*n* = 120); SCID, 102.5 (*n* = 24); NOD-SCID, 102.17 (*n* = 24); NOG, 101.1 (*n* = 75). It is notable that during the course of experiments covering 9 months of observation, we also observed spontaneous thymomas with a frequency of some 14% in NOD-SCID mice in agreement with previous reports [16], which makes interpretation of tumorigenicity tests with NOD-SCID mice complicated.

Based on the preceding data, we chose NOG mice for subcutaneous tumorigenicity testing of iPSC-derived RPE, assuming that NOG mice could generate tumors from the lowest number of residual PSC or tumorigenic non-terminally differentiated PSC-derived cells. We then subcutaneously transplanted various doses of iPSC (201B7, Riken CDB) with Matrigel® into NOG mice to determine TPD50 for iPSC. The TPD50 value for iPSC (201B7) via the subcutaneous route was 102.12 (*n* = 20) over 84 weeks of observation [9] (Figure 1). Tumorigenicity tests via a subcutaneous route with NOG mice is a sensitive quality control test to detect a small number of remaining PSC in PSC-derived investigational product regardless of cell type. Of course, the TPD50 for iPSC transplanted via a clinical route can be checked independently. In our case, we used nude rats for tumorigenicity testing via a clinical route, as the subretinal space of mice is very small and transplanting cells via a clinical route requires outstanding technique by a skilled operator. Thus, we needed larger animals to avoid "false" negative results due to failure of transplantation, to transplant a clinically relevant dose of GMP-grade iPSC-derived RPE (without Matrigel) and to confirm that the transplantation of brown colored RPE was in the right position in the albino eye ball of nude rats [9]. We did not use any "AMD" disease model animals [17,18] because they will not recapitulate all the features of human AMD. In human AMD, the macular region is focally affected and the rest of the retinal area is intact. Treatment of human wet-type AMD with an iPSC-derived RPE sheet is conducted by transplanting the RPE sheet into the affected lesion after removal of choroidal neovascularization. Thus, we assumed that a transplanted RPE sheet would receive a trans-effect from the intact retina. For that reason, we transplanted the RPE sheets into nude rats with intact retinal function rather the recapitulate the microenvironment of the clinical setting. Thus, the choice of animal should be made depending on the degree of immunodeficiency, anatomical demands and planned clinical manipulation. The TPD50 value for iPSC or HeLa cells via the clinical route was 104.74 (*n* = 26) or 101.32 (*n* = 37) respectively (Figure 2). The large discrepancy between the TPD50 values for iPSC and that of HeLa prompted us to examine the effect of the microenvironment on iPSC-derived products to better design tumorigenicity tests via the clinical route (see below).


**Figure 2.** Tumorigenicity test via clinical route with Nude rats. A table in above showed type of cells used as a positive control for tumorigenicity test (iPSC cell line 201B7 and tumor cell line HeLa), minimum dose for tumor formation and Log10 TPD50 for them when transplanted via clinical route. A line graph showed value for Log10 TPD50 for iPSC or HeLa at respective monitoring point (0–55 or 64 weeks). Photos (left from top to bottom); NC: non-transplanted control, iPSC: iPSC transplanted mouse. iPSC-transplanted (iPSC) or non-treated control (NC) eye ball. HE staining of slice section of iPSC-transplanted eye ball. Photos (right top to bottom) histology of teratoma formed; cartilage (mesoderm), intestinal tissue-like (endoderm) or neuron-like (ectoderm) tissue.

Another option to address the tumorigenic potential of autologous iPSC-derived products is to transplant rodent cells into a rodent with same genetic background to evade immune rejection associated with xeno-transplantation. Of course, it will be necessary to accumulate sufficient data to demonstrate that rodent cells used in this test are equivalent to human investigational cell products before starting the test.

#### **5. Administration Route and Microenvironment at the Transplantation Site**

The administration route should mimic the clinical route as closely as possible to address the tumorigenic potential of investigational cells in the context of the microenvironment at the transplantation site. Therefore, evaluation of the microenvironment of the transplantation site

including trans-effects from the microenvironment on investigational cells should be assessed prior to the commencement of large scale tumorigenicity testing. In the event of teratoma formation by residual undifferentiated PSCs, trans-effects of host tissue on PSC should be examined. Towards this end, we have established an *in vitro* co-culture system by placing PSC in culture inserts and culturing host or human primary tissue on the bottom of the dish. When iPSCs in culture inserts were co-cultured with cardiomyocytes or neural cells in the bottom of the dish, the growth of iPSC was not affected, but when they were co-cultured with RPE, the number of iPSCs was reduced drastically [19]. We found that RPE secreted Pigment Epithelium-derived Factor (PEDF). Addition of anti-PEDF antibody into the co-culture system blocked the reduction of iPSC cell number. Further addition of recombinant human PEDF (hrPEDF) induced apoptotic cell death and dramatically reduced ESC and iPSC cell number. hrPEDF did not show any reduction in the number of HeLa cells. Indeed, the TPD50 for iPSC was 104.75 when transplanting into the subretinal space (clinical route), while that for HeLa was 101.32. That means that approximately 20 HeLa cells could generate a tumor in the subretinal space in half of the rats transplanted, but more than 5 × 104 iPSCs were required to generate teratomas in the subretinal space in half of the rats transplanted. As we transplanted 0.8–1.5 × 104 iPSC-derived RPE cells in sheets via the clinical route in tumorigenicity tests, it is unlikely that we could observe teratomas from tumorigenicity tests via the clinical route. Further tests, such as transplanting serial dilutions of iPSC in the final product in the subretinal space would not be informative and cannot be justified if tried. However, tumorigenicity tests via the clinical route could be useful to address the tumorigenic potential of non-terminally differentiated tumorigenic cells in iPSC-derived RPE products. This test would be sensitive enough to detect tumors in half the rats transplanted with 20 HeLa cells. We conducted this test for this reason and observed no tumor-forming event (*n* = 36) during a 10–20 months monitoring period. The lack of tumor-forming events was eventually confirmed by IHC of transplanted cells in host tissue section.

We point out that the risk of teratoma formation by a small number of residual iPSC in iPSC-derived RPE in a clinical setting should be thoroughly addressed especially for autologous cell transplantation. Towards this end, subcutaneous tumorigenicity tests are being conducted concurrently with NOG mice wherein we transplant 1 × 106 cells embedded in Matrigel. This test is sensitive enough to detect as few as 10 iPSCs [8]. We have conducted this test with 71 animals that were monitored for 9 to 21 months and obtained negative result after examination of tissue sections by IHC.

In addition, we reported a highly sensitive residual hiPSC detection method based upon qRT-PCR using primers for the *LIN28A* transcript [20] in hiPSC-derived RPE. This method enabled us to detect residual hiPSCs down to 0.002% of differentiated RPE cells. These assays were effective quality control tests and test cells with negative results with this qRT-PCR test could be used for tumorigenicity testing and therapy. We conclude that even if a few (less than 10) autologous iPSCs are present in an iPSC-derived cell product, the chance of developing a teratoma is negligible when transplanted into the subretinal space.

#### **6. Monitoring Period**

We subcutaneously transplanted various doses of HeLa cells with or without Matrigel® into Nude, SCID, NDO-SCID and NOG mice and into the subretinal space of nude rats. We also subcutaneously transplanted various doses of iPSC with or without Matrigel® into NOG mice or into the subretinal space of NOG mice. As HeLa cells and iPSC can generate tumors in NOG mice with a relatively small number of cells, a long observation period can be required so that a tumorigenic event originating from a small number of transplanted cells is not overlooked. Ten HeLa cells needed 18 weeks and 10 iPSCs needed 40 weeks to generate tumors in NOG mice in the longest cases. Ten HeLa cells needed 33 weeks and 1 × 104 iPSCs required 33 weeks to generate tumors in the subretinal space of nude rats in the most protracted cases. Overall, it is recommended that the immunodeficient rodents be monitored up to 12 months so that a tumor formation event is not missed and to conduct satisfactory statistical analyses.

#### **7. Detection of Transplanted Cells**

Tumor formation by transplanted human cells can be detected regardless of cell type (teratoma or tumor) by staining tissue sections of the transplant site in host animal with human-specific antibody and Ki67. Nuclear staining with DAPI or Hoechst will not demonstrate that the cells in the tissue section were viable at the time of sacrifice, but sharp margins of the nuclear membrane will suggest that cells were alive and free from autophagy or necrotic events. Human-specific antibodies such as STEM121 (StemCells, AB-121-U-050), Lamin A + C (Abcam, AB108595), and HNA clone 3E1.3 Millipore MAB4383) can be used to identify human cells in host tissue. *In situ* hybridization with a species-specific (human, mouse, rat, *etc.*) probe may generate clear signals, but it may require elaborate sample preparation steps when a paraffin section is used. Tumor-forming cells with proliferation potentials were clearly distinguished by positive staining with Ki67 (MIB-1, Dako M7240) [9]. Further staining of human cells with antibodies specific for human differentiation markers will clearly identify the transplanted human cells.

#### **8. Dose, Number and Sex of Immune Deficient Animals**

The dose used in tumorigenicity tests should be determined in the context of the intended clinical use. In general, toxicology tests or POC tests require an escalation of doses to define the safety margin or the effective therapeutic margin. However, this may not be the case with tumorigenicity tests as they aim to address the tumorigenic potential of the maximum dose of the cell product that will be used in therapy. Considering the body size of the animal and anatomical space of the receptive transplant site in the animal, a relevant dose should be administered via the clinically route. In our case, we transplanted 0.8–1.5 × 104 iPSC-derived RPE cells into the subretinal space of nude rats and 1 × 10<sup>6</sup> iPSC-derived RPE cells with Matrigel® subcutaneously, based on the fact that we intended to transplant 4–8 × 104 iPSC-derived RPE in the clinic. We transplanted a maximum or supra-maximum test dose to minimize the risk of underestimating tumor-forming events in a clinical setting.

The number of rodents in each group should be more than 6 for statistical analysis to obtain significant results using the Clopper-Pearson method. If the cell therapy focuses on a single gender, the sex of mice should be matched in the tumorigenicity test. If not, female mice should be chosen to conduct the tests as stated in WHO TRS 878. Male mice attack cage mates, which leads to a reduction of animal number during long-term monitoring.

#### **9. Conclusions**

It is important to design animal tumorigenicity tests so that they do not underestimate the frequency of tumorigenic events in a clinical setting, based on risk assessment of the respective test. In this review, we have highlighted points to be considered by emphasizing the possible risks and the countermeasures we have taken against them. It is important to gather genetic information from the PSC-derived cell product by CGH array, mBAND and FISH analysis in a timely manner. We need to evaluate the effect of the microenvironment on test cells at the transplant site and the tumor-forming potential of test animals via both the clinical route and via the subcutaneous route. The latter would serve as a sensitive quality control test. This analysis must be mindful of the required dose, type and duration of monitoring and application of an effective IHC method to detect and evaluate the transplanted cells. Conducting pilot studies will help to obtain some of the information and design informative pivotal tests. Clinical researchers need to fully understand the scope and limit of each preclinical test to predict adverse events in the clinic.

#### **Acknowledgments**

We thank Chikako Morinaga of Laboratory for Retinal Regeneration, RIKEN Center for Developmental Biology for critical advising of the experiments, Yoji Sato of National Institute of Health Sciences, Tokyo for scientific discussion, Masayuki Shikamura of FBRI and Hiroyuki Kamao of Department of Ophthalmology, Kawasaki Medical School for rodent studies and Mamoru Ito of CIEA for supplying NOG mice.

#### **Author Contributions**

Conceived and designed the experiments: Hoshimi Kanemura, Masahiro J. Go and Shin Kawamata. Performed the experiments: Hoshimi Kanemura. Contributed reagents and materials/analysis tools: Noriko Sakai and Masayo Takahashi. Wrote the paper: Shin Kawamata, Hoshimi Kanemura and Masahiro J. Go.

#### **Conflict of Interests**

The authors declare no conflict of interest associated with this manuscript.

#### **References**

1. Strauss, S. Geron trial resumes, but standards for stem cell trials remain elusive. *Nat. Biotechnol.* **2010**, *28*, 989–990.


### **Concise Review: Methods and Cell Types Used to Generate Down Syndrome Induced Pluripotent Stem Cells**

#### **Youssef Hibaoui and Anis Feki**

**Abstract:** Down syndrome (DS, trisomy 21), is the most common viable chromosomal disorder, with an incidence of 1 in 800 live births. Its phenotypic characteristics include intellectual impairment and several other developmental abnormalities, for the majority of which the pathogenetic mechanisms remain unknown. Several models have been used to investigate the mechanisms by which the extra copy of chromosome 21 leads to the DS phenotype. In the last five years, several laboratories have been successful in reprogramming patient cells carrying the trisomy 21 anomaly into induced pluripotent stem cells, *i.e.*, T21-iPSCs. In this review, we summarize the different T21-iPSCs that have been generated with a particular interest in the technical procedures and the somatic cell types used for the reprogramming.

Reprinted from *J. Clin. Med*. Cite as: Youssef Hibaoui and Anis Feki. Concise Review: Methods and Cell Types Used to Generate Down Syndrome Induced Pluripotent Stem Cells. *J. Clin. Med.* **2015**, *4*, 696–714.

#### **1. Introduction**

Down syndrome (DS), caused by a trisomy of chromosome 21 (HSA21), is the most common genetic developmental disorder, with an incidence of 1 in 800 live births. DS individuals show cognitive impairment, learning and memory deficits, arrest of neurogenesis and synaptogenesis, and early onset of Alzheimer's disease [1,2]. They are also at greater risk of developing acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML). The incidence of ALL, the most common leukemia in childhood, is approximately 20-fold higher in children with DS than in the general population. The incidence of AML is between 46- to 83-fold higher, with a particular susceptibility to acute megakaryoblastic leukemia [3]. The detailed pathogenetic mechanisms by which the extra copy of HSA21 leads to the DS phenotype remain unknown. However, there is evidence that several regions exist on HSA21 with various "dosage sensitive" genes contributing to a given phenotype, which could also be modified by other genes on HSA21 and in the rest of the genome [4,5].

Several models have been used to recapitulate the DS phenotype, such as mouse models [6]. However, they do not accurately recapitulate the specificities of the human phenotype. A new finding indicating that induced pluripotent stem cells (iPSCs) can be reprogrammed through the introduction of a few factors [7,8] has opened a new avenue for the investigation of neurological diseases (reviewed in [9]). The first application of this technology appeared only one year after the release of these articles, with the derivation of iPSC lines from patients affected by several diseases including trisomy 21 [10]. Since that research paper, a dozen other studies reporting the generation of trisomy 21 iPSCs (T21-iPSCs) have appeared in the last five years. In this concise review, we will

summarize the T21-iPSCs that have been reported up to now with a particular focus on the origin of the somatic cells and the procedures used for the reprogramming.

#### **2. Procedures Used for the Reprogramming of T21-iPSCs**

Direct reprogramming into iPSCs involves the ectopic introduction of a set of core pluripotency-related transcription factors in a somatic cell. In the vast majority of iPSC studies, *OCT4* (also known as *POU5F1*), *SOX2*, *KLF4* and *MYC* (also known as *c-MYC*) are used for the reprogramming into pluripotency as in the original study by Yamanaka's team [7]. In addition to this so-called OSKM cocktail, Thomson and colleagues also proposed another reprogramming cocktail that comprises *OCT4* and *SOX2* but *NANOG* and *LIN28* instead of *KLF4* and *c-MYC*: the so-called OSNL cocktail [8]. When this process is successful, compacted colonies appeared in the culture dish that showed marked similarities to embryonic stem cells (ESCs) with respect to morphology, growth properties, expression of pluripotency factors, self-renewal and developmental potential [7,8,11]. The current published T21-iPSC lines have been all generated with the OSKM cocktail, except for one study where T21-iPSCs were derived with the OSNL cocktail [12]. Thus, these T21-iPSC lines were derived predominantly through integrative delivery systems and, to a lesser extent, through non-integrative delivery systems (Table 1).


**Table 1.** The different T21-iPSCs reprogrammed.


**Table 1.** *Cont.*

DS: Down syndrome; iPSCs: induced pluripotent stem cells; reprogramming cocktails: O for OCT4, S for *SOX2*, K for *KLF4*, M for *c-MYC*, N for *NANOG*, L for *LIN28*.

#### *2.1. Integrative Procedures Used for the Derivation of T21-iPSCs*

The first T21-iPSC lines were generated with the OSKM cocktail using the Maloney murine leukemia virus (MMLV)-derived retroviruses pMXs [10]. MMLV-derived retroviruses have been used in more than half of the studies reporting the generation of T21-iPSCs (Table 1). In this respect, MMLV-derived retroviruses allow the delivery of genes into the genomes of dividing cells, and the efficiency of iPSC generation from human fibroblasts using MMLV-derived retroviruses is approximately 0.01%.

Lentiviral vectors have also been successfully used to reprogram T21-iPSCs (Table 1). They are generally derived from HIV. They exhibit higher infection efficiency than MMLV-derived retroviruses and allow the delivery of genes into the genome of dividing and non-dividing cells. The efficiency of iPSC generation from human fibroblasts using lentiviral vectors is comparable to those of MMLV-derived retroviruses (~0.01%). However, compared to MMLV-derived retroviruses, lentiviruses are less repressed in human pluripotent stem cells (hPSCs) [26]. In this respect, a major improvement has been seen in the method with the development of single polycistronic vectors containing all the reprogramming factors, which reduce multiple transgene insertion into the genome [27]. Moreover, in one study, T21-iPSCs were derived through doxycycline-induced lentiviral vectors with an OSKM cocktail [18]. The main advantage of this method is that it allows greater control over transgene expression; compared with constitutive lentivirus, in which the vector is integrated and then may or may not be silenced, the doxycycline-induced lentivirus is integrated and silenced when doxycycline is removed. A more recent improvement of the method has been the introduction of lentiviral vectors that incorporate loxP sites allowing their excision via Cre recombinase when pluripotency is achieved [28]. However, viral elements flanking the loxP sites still remain after excision.

The use of integrating vectors offers a more efficient means of reprogramming but also raises major drawbacks with the risk of (i) genetic and epigenetic aberrations; (ii) overexpression of potentially tumorigenic genes such as *c-MYC*; and (iii) incomplete silencing of reprogramming factors following differentiation. Also, the use of integrative approaches has been associated with genomic instability of the generated iPSCs. Genomic instability in iPSCs could come from various sources, which means karyotype analysis is one of the first verifications that has to be done when establishing an iPSC-based disease model. Mutations can originate from the parental somatic cells from which the iPSCs are derived or can be generated during the reprogramming process [29]. However, this is still debated, as growing evidence supports a similar frequency of genetic aberrations in iPSCs, independently of the reprogramming method (integrative or non-integrative) or the cell type used for the reprogramming [29–33]. Alternatively, it could be acquired after culture adaptation and passaging over time [34,35]. For example, mechanical passaging appears to produce more stable cells with a normal karyotype than enzymatic harvesting methods [36–38]. This genomic instability is not restricted to long-term culture, but can appear very rapidly, within five passages after switching human ESCs to enzymatic dissociation [39].

Another major concern of integrative delivery systems is related to a possible transgene reactivation that could lead to the overexpression of potentially tumorigenic genes such as *c-MYC* or *KLF4*. For instance, the presence of *c-MYC* is a major limitation, as chimeras derived from iPSCs frequently develop tumours due to the reactivation of *c-MYC* [40,41]. Therefore, transgene silencing has to be investigated after initial expansion of a few passages of the newly generated iPSCs. Moreover, early reports have proposed that residual transgene expression (of *c-MYC* or *KLF4* in particular), after using integrating viral approaches may affect pluripotency and differentiation states [8,11]. It is important to note, however, that reprogramming approaches that exclude *c-MYC* are more labor-intensive and less efficient. In fact, *c-MYC* is an important inducer of reprogramming [42–45], activating pluripotent genes and maintaining the pluripotent state of PSCs [46–48]. It is considered the driver of the first transcriptional wave during cellular reprogramming into iPSCs [49]. This could explain, at least in part, why the vast majority of the reported iPSC lines are achieved using *c-MYC*. Of note, other potential contributors of tumorigenicity of iPSCs have been reported; in particular, we highlighted the crucial role of *NANOG*  during reprogramming into iPSCs with respect to germ cell tumor formation [50].

Regarding the impact of these methods on the differentiation potential of iPSC lines, Hu *et al.* reported variable potency of iPSCs to differentiate into neural cells independently of the set of reprogramming transgenes used to derive iPSCs as well as the presence or absence of the reprogramming transgenes in the generated iPSCs [51]. In line with this, in a study comparing the differentiation potential of iPSC lines derived from a single parental fibroblast line via several reprogramming strategies (+/í *c-MYC*, excised or non-excised transgene), neither the presence of *c-MYC* nor the presence of the transgene removed the *in vitro* potential of these iPSCs to differentiate into neuroprogenitor cells, neurons, astrocytes and oligodendrocytes [52]. Furthermore, it appears that omission in iPSCs of reprogramming factors, and of *c-MYC* in particular, compromises the efficiency of their subsequent differentiation into neuroprogenitor cells and neurons [53].

#### *2.2. Non-Integrative Procedures Used for the Derivation of T21-iPSCs*

Two non-integrative approaches have been used for the generation of T21-iPSCs: episomal vectors [19] and Sendai virus vectors [20]. Briggs *et al.* reported the first generation of T21-iPSCs free of vectors and transgenes [19]. This reprogramming was achieved by transfection with oriP/Epstein-Barr nuclear antigen-1 (oriP/EBNA1)-based episomal vectors [54]. These plasmids can be transfected without the need for viral delivery and can be removed from cells by culturing in the absence of selection. In other terms, the exogenous DNA is not integrated into the iPSC genome. However, the reprogramming efficiency of this approach for human fibroblasts is extremely low, ~0.0006% [54].

An alternative non-integrative method has been used for the generation of T21-iPSCs by the mean of Sendai virus [20]. Sendai virus, a member of the Paramyxovirus family is an enveloped virus with a nonsegmented negative-strand RNA genome. Modified Sendai virus (through the deletion in one of the two envelope glycoproteins) has emerged as an efficient and robust RNA-based gene delivery system. Since Sendai virus RNA replication occurs in cytoplasm of the infected cells without a DNA phase, there is no risk of vector genome integration into host genome [55]. Thus, the efficiency reached by this method is much higher than that achieved with episomal vectors for the reprogramming of human fibroblasts to iPSCs: ~1% [55].

#### **3. Age and Type of the Donor Cells Used for the Reprogramming**

Reprogramming into iPSCs requires the delivery of pluripotency factors into a somatic cell. This is achieved with different efficiencies and kinetics depending on the donor cell type. Therefore, the choice of the type of the donor cells is an important aspect to consider before the generation of disease-specific iPSCs. As for 80% of the studies reporting the derivation of human iPSCs, fibroblasts remain the cell type the most commonly used for the derivation of T21-iPSCs (Table 1). There are many reasons for this. Even though dermal fibroblasts are obtained from skin biopsies or neonatal foreskin biopsies, which require invasive procedures, they present several advantages. First, the culture of fibroblasts is relatively easy and cheap. In culture, fibroblasts also exhibit a high proliferation rate, viability and stability (at least in low passages, as the risk of accumulated genomic alteration increases with passaging). Moreover, the discovery of iPSC technology has been done initially in mouse fibroblasts [56] and subsequently adapted in human fibroblasts [7,8]. Then, most of the data available on the relative kinetics and efficiencies of the different methods used for the reprogramming have been characterized using fibroblasts as donor's cells (reviewed in [57]). In line with this, most of the iPSCs banked have been generated with fibroblasts as a starting material. All these considerations make fibroblasts as the main cell type used for the reprogramming in general as well as in DS research. However, other cell type has been used for the generation of T21-iPSCs such as cells from amniotic fluids which are more easily obtained and reprogrammed into iPSCs [16]. Indeed, second semester amniocenteses are routinely collected in the context of prenatal diagnosis screening. Also, compared with fibroblasts, cells from amniotic fluids transduced with OSKM exhibited higher efficiency (100 times more) and are reprogrammed into pluripotency more than twofold faster [58]. This makes cells from amniotic fluids as easy to reprogram as keratinocytes [59]. Similarly, fetal stromal cells and mononuclear cells have been used for the generation of T21-iPSCs [18].

During the reprogramming process, the epigenetic state of the donor's cells has to be reset to obtain a pluripotent state; this includes modification of the DNA methylation profile, and chromatine marks [60,61]. However, genome wide DNA methylation studies showed that iPSCs retain the DNA methylation signature of the donor's cells [60,62]. This so-called "epigenetic memory" consists of residual specific marks of the parental somatic cells that escape the reprogramming process, leading to a preferential differentiation potential of the generated iPSCs into the tissue of origin rather than other lineages [60,61]. For instance, iPSCs derived from cord blood display a higher capacity for hematopoietic differentiation than iPSCs derived from keratinocyte, and reciprocally [60]. However, it is important to note that studies investigating donor epigenetic memory of iPSCs have confounded the donor's cell type and the donor genetic background due to the practical difficulty of collecting various primary tissues from the same donor. Also, it has been reported that donor epigenetic memory appears to be gradually lost after prolonged iPSC culture [60,62,63], which supports the idea that the preferential differentiation potential due to epigenetic memory can be overcome. Moreover, there are some indications that non-coding RNAs such as miRNAs play a role in maintaining residual memory of donor cells in iPSC-derived cells [64,65]. For instance, miR-155 have been identified as a key player in somatic donor memory of iPSCs in the context of iPSC differentiation toward hematopoietic progenitors [64].

Another important factor that should be considered when deriving disease specific iPSCs is the age of the donor's cells. T21-iPSCs have been generated from DS tissue from fetal, neonatal and adult stages (Table 1). In this respect, embryonic tissue appears to be more prone to reprogramming into pluripotency than adult tissue. Barriers such as the age and the differentiation status of the donor's cells could explain this property [66–68]. For instance, it has been shown that the increased levels of the age-related genes *p16* (*INK4A*), *p19* (*ARF*) and *p15* (*INK4B*), which encodes two tumor suppressors, limit the efficiency and the fidelity of the reprogramming [67]. Also, the differentiation stage of the starting cell used for the reprogramming has a critical impact on the efficiency of reprogramming into iPSCs. Blood progenitors reprogram into iPSCs up to 300 times more efficiently than terminally differentiated blood cells [68]. Similarly, neural progenitor cells which express *SOX2* endogenously have only been successfully reprogrammed into iPSCs with *OCT4* [69]. Considering that donor cell type and age may affect the differentiation potential of the iPSCs, it is crucial to establish D21-iPSCs and T21-iPSCs from the same parental somatic cells at the same developmental age.

#### **4. Isogenic D21-iPSCs and T21-iPSCs**

Among the potential variables that must be considered when establishing an hPSC-based disease model, the definition of a non-disease control is of crucial importance [70,71]. The genetic background of both control and the affected cells has to be identical or similar in order to be sure that the differences observed in the studies are due only to the disease and not to the choice of either the control or the affected samples. Traditionally, iPSCs from unrelated healthy individuals together with ones from age-matched, unrelated affected patients are often used to decrease the variability of individual genetic background and the variability among the iPSC lines regarding their *in vitro* differentiation potential. To overcome these problems, several approaches have been developed to obtain isogenic D21-iPSCs and T21-iPSCs. This is particularly important as isogenic D21-iPSCs and T21-iPSCs represent an ideal situation for the investigation of the effect of the supernumerary HSA21 on the DS phenotype, since the rest of the genome is theoretically identical. It could also limit the need to generate several iPSC lines.

Chromosomal aberrations have been often observed after culture adaptation over time in hPSCs [34]. In particular, stable genomic aberrations that confer growth, self-renewal, and differentiation advantages for hPSCs are often selected over time [29,34,72]. In the study by MacLean *et al*., one clone of T21-iPSCs lost one copy of HSA21 with culture passages leading to a mixed culture of isogenic D21-iPSCs and T21-iPSCs. Then, they succeeded in isolating isogenic D21-iPSCs and T21-iPSCs from this mixed culture by cultivating them as single cells and discriminating D21-iPSCs from T21-iPSCs by FISH analysis (Figure 1A) [17]. This event seemed to occur also for one clone of T21-iPSCs generated by Chen *et al*. [25].

In another study, Li *et al.* succeeded in deriving isogenic D21-iPSCs from T21-iPSCs. For this, they used an adeno-associated virus to introduce a *TKNEO* transgene into one copy of HSA21 of T21-iPSCs. When the T21-iPSCs were grown in a medium that selected against *TKNEO*, the only cells that survived were the ones that spontaneously lost the extra HSA21 (Figure 1B) [12].

In an elegant study, Lawrence *et al.* have shown that the extra copy of HSA21 in T21-iPSCs can be silenced through the insertion of the RNA gene called *XIST*, a gene responsible for the silencing of one of the two X-chromosomes in female cells. Interestingly, they demonstrated that the insertion of *XIST* gene at a specified location in the HSA21 using zinc finger nuclease technology effectively repressed genes across the supernumerary HSA21 in T21-iPSCs, leading to the generation of isogenic D21-iPSCs and T21-iPSCs (Figure 1B) [21].

It is well known that varying degrees of mosaicism for trisomy 21 may exist in the generation population; it represents 1%–3% of DS cases [73]. This leads to a combination of euploid cells and cells carrying trisomy 21 anomaly within individual tissues (reviewed in [74]). Taking advantage of this rare situation, two recent studies reported the derivation of isogenic D21-iPSCs and T21-iPSCs from fibroblasts from an individual mosaic for trisomy 21 (Figure 2A) [19,20].

**Figure 1.** Isogenic iPSCs obtained through spontaneous or induced loss of trisomy 21. Isogenic D21-iPSCs and T21-iPSCs have been obtained either via spontaneous or induced loss of one copy of HSA21. (**A**) T21-iPSCs can lose one copy of HSA21 after culture adaptation and passaging over time [17,25]; (**B**) The loss of one copy of HSA21 in T21-iPSCs has been induced through the insertion of a foreign gene called *TKNEO* into one copy of HSA21 (within the *APP* gene) of T21-iPSCs. When these T21-iPSCs were grown in a medium that selected against *TKNEO*, the most common reason for the cells to survive was the loss of one copy of HSA21 [12]. The silencing of one copy of HSA21 in T21-iPSCs has been induced through the insertion of *XIST* into one copy of HSA21 of T21-iPSCs. This leads ultimately to the generation of isogenic D21-iPSCs [21].

Most monozygotic twins are "genetically identical" and are in general expected to be concordant for health, chromosomal abnormalities, and Mendelian disorders. However, in very rare cases, monozygotic twins can be discordant for the disease (reviewed in [75]). One example of this is monozygotic twins discordant for trisomy 21 [76]. We exploited this rare and unique situation by deriving iPSCs from fetal fibroblasts of monozygotic twins discordant for trisomy 21 [22–24] and thus confounding effects from genomic variability were theoretically eliminated (Figure 2B).

**Figure 2.** Isogenic iPSCs from individual mosaic for trisomy 21 or from monozygotic twins discordant for trisomy 21. (**A**) Isogenic D21-iPSCs and T21-iPSCs have been derived from mosaic patients for trisomy 21 [19,20]; (**B**) Isogenic D21-iPSCs and T21-iPSCs have been generated from monozygotic twins discordant for trisomy 21 [22–24].

#### **5. Down Syndrome Phenotype Investigated**

Among the phenotypes observed in DS individuals, only two have been explored using T21-iPSCs, namely brain-related defects and myeloid leukemia.

#### *5.1. Brain-Related Defects*

Five groups, including our own, have reported the recapitulation of the relevant DS phenotype using neurons derived from T21-iPSCs. Consistent with a DS post-mortem human brain, T21-iPSCs showed reduced neurogenesis when induced to differentiate into neuroprogenitor cells (NPCs) and further mature into neurons [19,21,23]. This effect was associated with a proliferation deficit and increased apoptosis of NPCs derived from T21-iPSCs [23]. Thus, together with the reduced neurogenesis, T21-iPSCs showed a greater propensity to generate both astroglial [19,23] and oligodendroglial cells [23] upon neural induction and differentiation. This gliogenic shift appeared early in development as it starts at the NPC level [23]. Moreover, neurons derived from T21-iPSCs exhibited not only a reduction of their population but also structural alterations compared to those derived from D21-iPSCs. They exhibited in particular reduced dendritic development [23] and reduced expression of synaptic proteins such as synapsin or SNAP25 [20,23]. In line with this, we found a lower proportion of excitatory glutamatergic synapses whereas the proportion of inhibitory GABA-ergic synapses was not substantially altered in neurons derived from T21-iPSCs [23]. Regarding the electrophysiological properties, neurons derived from T21-iPSCs displayed a significant synaptic deficit that affects excitatory glutamatergic synapses and inhibitory GABA-ergic synapses equally [20].

Furthermore, the increased proportion of astroglial cells at the expense of neurons upon neural induction and differentiation of T21-iPSCs [19,23] is of special interest as it has been shown that astrocytes derived from T21-iPSCs exhibited higher levels of reactive oxygen species (ROS) and lower levels of synaptogenic molecules than astrocytes derived from D21-iPSCs. This ultimately contributes to oxidative stress-mediated cell death and abnormal maturation of neurons derived from T21-iPSCs [25].

Finally, Shi *et al.* used T21-iPSCs as a PSC model of Alzheimer's disease pathology, given that DS individuals present early onset of Alzheimer's disease. They showed that cortical neurons derived from T21-iPSCs exhibited greater secretion of amyloid peptides, tau protein phosphorylation and cell death, supporting the notion that T21-iPSCs are an excellent model for AD study [13].

#### *5.2. Myeloid Leukemia*

Two recent studies have explored the potential of T21-iPSCs to model hematopoietic defects associated with trisomy 21 [17,18]. Using a differentiation protocol that mainly drives hPSCs towards primitive yolk sac-type hematopoietic progenitors, Chou *et al.* showed that hematopoietic progenitors derived from T21-iPSCs exhibit an increased propensity for erythropoiesis [18], similar to what it is observed in DS fetal liver hematopoiesis [77,78]. However, in contrast with DS fetal liver hematopoiesis, no difference was found between D21-iPSCs and T21-iPSCs in their capacity to generate megakaryocytes [18]. In the second study, MacLean and colleagues used a differentiation protocol that drives hPSCs towards definitive fetal-liver type progenitors. They found that hematopoietic progenitors derived from T21-iPSCs (and from T21-ESCs) exhibit higher multi-lineage colony-forming potential [17]. In particular, T21-iPSC-derived hematopoietic progenitors showed a greater colony-forming unit for erythroid, myeloid and megakaryocyte lineages [17], consistent with DS fetal liver hematopoiesis [77,78]. This indicates that trisomy 21 favours the expansion of hematopoietic progenitor cells. Altogether, these two studies point to different defects in primitive yolk sac-type hematopoietic progenitors and definitive fetal-liver type progenitors derived from T21-iPSCs and further suggest that the effects of trisomy 21 are likely specific to the developmental stages of the hematopoietic progenitors. Further studies using this iPSC-based model should provide important clues regarding the impact of trisomy 21 on hematopoietic development.

#### **6. Conclusions and Perspectives**

Since the first paper demonstrating that fibroblasts from DS patients can be reprogrammed into iPSCs by retroviral delivery of OSKM cocktail [10], several alternative methods and cell types have been used to generate T21-iPSCs (Table 1). At the moment, there is no consensus for the cell type that should be used for the reprogramming. The choice of the starting material depends not only on the availability of the cell type, but also on the ability and efficiency of these cells for reprogramming. With respect to the reprogramming method that should be used, this depends mostly on the priorities regarding the applications of the generated iPSCs. The priorities are not the same if the generated iPSCs aimed at investigating (i) the reprogramming mechanisms; (ii) disease modelling and drug screening and (iii) regenerative medicine. For the former aim, as the reprogramming approach needs to be efficient, the integrative inducible lentiviruses will meet most of the requirements. The safety of the generated iPSCs is a major requirement for clinical applications but less crucial for disease modelling and drug screening studies. In this respect, Sendai viruses and mRNA methods offer the advantage of generating iPSCs free of vectors and transgenes with a high efficiency [79].

Another major concern when generating iPSCs is the definition of a non-diseased control. In most of the studies reporting disease modelling using iPSCs, iPSC lines from unrelated healthy donors have been used as controls since genetically matched non-diseased controls are often difficult to obtain. In this respect, isogenic D21-iPSCs and T21-iPSCs offer the unique opportunity to study the effect of the supernumerary HSA21 on DS phenotype without the biological "noise" that could result from the variability of individual genetic background. These isogenic D21-iPSCs and T21-iPSCs has been achieved via several ways: (i) by spontaneous or induced loss of one copy of HSA21 in T21-iPSCs [12,17,21,25]; (ii) isogenic D21-iPSCs and T21-iPSCs from an individual mosaic for trisomy 21 [19,20]; (iii) isogenic D21-iPSCs and T21-iPSCs from monozygotic twins discordant for trisomy 21 [22–24]. Of note is the recent progress in genomic editing technologies such as transcription Activator-Like Effector Nucleases (TALEN), Zinc Finger Nucleases (ZFN) and Clustered Regularly Interspaced Short Palindromic Repeats (CRISPRs) (for review [80]) should provide opportunities to investigate genotype-phenotype correlations using "gene-edited" iPSC lines. For instance, it should allow the study of the contribution of candidate genes on DS phenotype by the investigation of the effect of genetic loss-of-function in T21-iPSCs and gain-of-function in D21-iPSCs of HSA21 genes in the target cell type of interest for DS.

A major drawback of iPSC technology is the variability that can appear at each step of the reprogramming and the differentiation processes. Reprogramming into iPSCs can give rise to unpredictable alterations of the genome such as copy number variants, karyotypic abnormalities, point mutations and deletions, epigenetic memory of the parental somatic cells [29–39,60–63]. Therefore, it is possible that such genetic and epigenetic alterations can affect the fidelity of the results regarding disease modeling and drug screening. Also, there is evidence that iPSC lines display variable potency to differentiate into the cell type of interest [51,60]. However, it is unclear what factors contribute to this variable efficiency of the iPSC differentiation, as it appears independent of the methods used for the reprogramming [51]. For this reasons, it is important to generate several iPSC lines from accurately chosen tissue of multiple normal and DS individuals, using them in priority non-integrative procedures. Such efforts will improve the identification of the pathogenetic mechanisms involved in DS by reducing the noise that could result from the variability of individual genetic background and from the experimental artifacts. At the same time, it will reduce the discovery of false pathogenetic mechanisms.

Another aspect that should be taken into account in DS modelling using iPSCs is the presence of a broad phenotypic variability among DS individuals. Even though DS individuals share some morphogenetic characteristics [1,4,5], trisomy 21 can have differential pathogenicity on individual genomes [81]. For example, brain-related defects are common traits in all DS individuals but other traits such as congenital heart defects only occur in ~40% of them. In line with this, cases of partial trisomy 21 and other HSA21 rearrangements associated with DS features have been reported [4,5]. Such cases could serve to link genomic regions of HSA21 with specific phenotypes given the possibility of generating the target cell type of interest for DS using T21-iPSCs.

Regarding the applications of T21-iPSCs, the abundance of studies reporting the generation of T21-iPSCs clearly shows that T21-iPSCs are reliable tool for DS modelling, given that the protocols for differentiation of iPSCs into neurons or hematopoietic cells are available. These protocols enable the production of large quantities of the target cell type for DS modelling. Some of these studies have been successful in recapitulating DS phenotypes using iPSCs (see Table 1). In this respect, transcriptional profiling of T21-iPSCs has proven extremely informative for the study of the pathogenetic mechanisms involved in DS phenotype [17–19,22–24]. For example, T21-iPSCs recapitulate the developmental disease transcriptional signature of DS [22–24]. Furthermore, T21-iPSCs allow the possibility of linking the genetic data to biological insights by deciphering the molecular changes in the target cell type of interest for DS (reviewed [82]). Then, the causal involvement of candidate HSA21 genes and pathways can be assayed by studies involving genetic loss-of-function in T21-iPSCs and gain-of-function in D21-iPSCs through genomic editing methods (for review [80]). Regarding DS modelling, only two phenotypes have been investigated so far: brain-related defects and myeloid leukemia (Table 1). However, other phenotypes associated with DS deserve investigations (heart defects, lymphoid leukemia and others). Moreover, modelling DS using iPSCs offers opportunities for drug screening. In concert with functional genomics, iPSCs form a powerful cellular model platform for drug screening assays with direct relevance to the DS phenotype. Integrating the genetic findings and the functional insights obtained from T21-iPSC-derived cells should provide a path to predict which drug might best counteract DS phenotype. Four studies have produced the proof of concept of such an application. Several proteins or pathways have been targeted and demonstrated beneficial effects on the DS phenotype, including oxidative stress-mediated cell death (with *N*-acetylcysteine, an antioxidant) [19], neurogenesis impairment (with epigallocatechine gallate, a DYRK1A inhibitor) [23], the gliogenic shift (with monocycline, an anti-inflammatory drug) [25] and AD-related phenotype (with inhibitors of gamma secretase) [13]. Finally, one promising aspect of iPSC technology is the potential use of these cells in cell replacement therapy to treat neurological diseases [9]. However, iPSCs have not been used until recently for clinical applications due to concerns over the immunogenicity and tumorigenicity of these cells [83,84]. Recently, iPSC technology has generated enthusiasm in the field of cell replacement therapy with the decision of Takahashi's team to treat a patient with a degenerative eye disease [85]. The possibility to induce the loss of one copy of HSA21 in T21-iPSCs and to produce subsequently isogenic D21-cells offers great hope for the treatment of some DS phenotypes (such as brain-related defects). However, numerous challenges remain for cell replacement therapy [9,86], and further studies are needed to address to which extent cells derived from iPSCs can be used for DS therapy. The coming years will tell whether these cells fulfil their potential.

In conclusion, we believe that T21-iPSC-derived cells are an invaluable resource for medical research. They will advance our understanding of the pathogenetic mechanism by which the extra copy of HSA21 leads to the DS phenotype. They have already offered the first opportunity to study the developmental events in the cell type of interest for DS: brain-related defects using iPSC-derived neurons and leukemia using iPSC-derived hematopoietic cells. IPSCs could also serve as a cellular platform for the evaluation of potential therapeutics.

#### **Acknowledgments**

This work was supported by grants from Dubois-Ferrière Dinu Lipatti, Gertrude Von Meissner and Novartis Foundations. The authors would like to especially thank Iwona Grad for useful comments and proofreading.

#### **Author Contributions**

Youssef Hibaoui and Anis Feki contributed to manuscript writing. Youssef Hibaoui prepared the figures and tables. Youssef Hibaoui and Anis Feki edited and revised the final manuscript.

#### **Conflicts of Interest**

The authors declare no conflict of interest.

#### **References**


Chapter 10: Immune Response

### **The Possible Future Roles for iPSC-Derived Therapy for Autoimmune Diseases**

#### **Meilyn Hew, Kevin O'Connor, Michael J. Edel and Michaela Lucas**

**Abstract:** The ability to generate inducible pluripotent stem cells (iPSCs) and the potential for their use in treatment of human disease is of immense interest. Autoimmune diseases, with their limited treatment choices are a potential target for the clinical application of stem cell and iPSC technology. IPSCs provide three potential ways of treating autoimmune disease; (i) providing pure replacement of lost cells (immuno-reconstitution); (ii) through immune-modulation of the disease process *in vivo*; and (iii) for the purposes of disease modeling *in vitro*. In this review, we will use examples of systemic, system-specific and organ-specific autoimmunity to explore the potential applications of iPSCs for treatment of autoimmune diseases and review the evidence of iPSC technology in auto-immunity to date.

Reprinted from *J. Clin. Med*. Cite as: Meilyn Hew, Kevin O'Connor, Michael J. Edel and Michaela Lucas. The Possible Future Roles for iPSC-Derived Therapy for Autoimmune Diseases. *J. Clin. Med.* **2015**, *4*, 1193–1206.

#### **1. Introduction**

Pluripotent stem cells have the ability to differentiate into all three of the embryonic germ layers, endoderm, mesoderm, or ectoderm. While these pluripotent cells may be of embryonic origin, somatic cells can be induced into this pluripotency state by transient ectopic expression of defined groups of transcription factors, hence the term "inducible" pluripotent stem cells (iPSCs). The advantages of inducing pluripotency includes the potential generation of unlimited numbers of required cells, deriving cells from hard-to-source tissues, reproduction of disease models, bypassing the ethical concerns regarding the use of embryonic stem cells and importantly provide an autologous cell therapy strategy that removes the need for immune suppression drugs.

#### **2. Background**

Following the seminal paper by Takahashi and Yamanaka [1], which reported using appropriate transcription factors, Oct4, Sox2, Klf4, and c-Myc, mouse fibroblasts could be reprogrammed into a pluripotent state, it has been demonstrated in human somatic cells. Furthermore, other combinations of transcription factors are able to induce pluripotency in human somatic cells as well [1–5].

Autoimmune diseases affect individual organs or a combination of organs, including the kidneys, brain, bone marrow, joints, or skin, however, the pathogenesis of most autoimmune diseases remains, at best, only partially delineated. IPSC technology has the potential to provide key cellular subsets which, given to patients, may alter their disease course by providing pure replacement of lost cells, may limit damage through immune-modulation of the disease process *in vivo*, and may provide substrates for the purposes of disease modeling *in vitro*. In this review, systemic lupus erythematosus (SLE) is taken as a prototypical example of a systemic auto-immune disease, along with rheumatoid arthritis (RA); diabetes mellitus (DM) as an example of organ specific autoimmunity; and multiple sclerosis (MS) as an example of system-specific neurological autoimmunity, to demonstrate the promising future research potential towards translational medicine of iPSC-derived treatment in a range of different contexts within Clinical Immunology.

#### **3. Disease Immunomodulation and Potential Cellular Components—SLE and RA as Examples**

The loss of tolerance to self is the fundamental basis of autoimmunity, with resultant aberrant immune responses of autoantibody formation and/or cellular immunity against self-tissue.

Systemic lupus erythematosus (SLE) is a prototypical systemic autoimmune disease. Usually affecting women of childbearing age, it is characterised by the production of multiple auto-antibodies directed against double-stranded DNA and other nuclear antigens, which are widely distributed throughout the body. The autoantibodies are produced by activated auto-reactive B cells following presentation of these self-antigens to self-reactive T cells. Along with autoantibody production are reduced populations of regulatory T cells (Tregs), reduced responses to regulation by these cells on effector T cells, immunological dysregulation and increased inflammation [6], immune complex formation and deposition, and end-organ damage, particularly if the disease affects the kidneys or central nervous system.

Rheumatoid arthritis is a symmetrical, inflammatory disease of synovial joints which also manifests extra-articular pathology in about 40% of patients. Affecting other parts of the musculoskeletal system, as well as the skin, eye, lung, heart, kidney, and vascular and nervous system tissues, it is likely that the inflammatory processes driving the synovial inflammation are also responsible for these extra-articular manifestations. RA patients develop autoantibodies to post-translationally modified synovial or stress-related proteins, which results in the conversion of arginine residues into citrulline (a process known as citrullination). In genetically susceptible individuals, preferential binding of these citrullinated self-peptides to MHC molecules may enable presentation to peripheral T cells, allowing expansion of potentially self-reactive T-cell populations. At the same time, if there is no presentation centrally in the thymus, there is no deletion or negative selection of autoreactive T cell populations, which is a possible mechanism for loss of self-tolerance in RA.

The mainstay of treatment, for both SLE and RA, is with immunosuppressive medications, however, true immunomodulation in the absence of toxicity is difficult to achieve.

There are a number of important cell populations that impact on systemic autoimmune disease course in which iPSC technology could potentially assist to model their effects and ideally contribute to regaining self-tolerance, such as regulatory T cells (Tregs) and dendritic cells. Targeting of particular cell lineages, rather than their end products, is also likely to be beneficial in the treatment of other autoimmunity diseases.

#### *3.1. Regulatory T Cells (Tregs)*

Regulatory T cells (Tregs), have an important role in the state of equilibrium that is immune tolerance, and are, therefore, also known as tolerogenic T cells. Tregs are CD4, CD25, and Foxp3 positive, and act to restrict the extent and duration of T cell mediated immune responses, and maintain peripheral tolerance by suppressing auto-reactive T cells that have escaped negative selection in the thymus. The mechanisms by which Tregs work continue to be discovered [7]. Most Tregs arise centrally in the thymus where cell lineage commitment is determined by T-cell receptor (TCR) specificity to self antigen. The transcription factor Foxp3 stabilises gene expression that specifies Treg differentiation while other transcription factors, including c-Rel, links TCR engagement and Foxp3 expression, within an appropriate cytokine and co-stimulatory molecule milieu, for Treg differentiation.

In the periphery, Tregs can be induced following repeated antigen exposure [8] under the influence of TGF-beta, converting Foxp3 negative T cells into Foxp3 positive induced Tregs (iTregs). Hence, this replaces T effector populations with regulatory populations, converting harmful responses to beneficial regulatory responses.

The list of potential defects in Tregs leading to autoimmune diseases are many (Table 1). Considering this extensive list, however, enables multiple potential targets for iPSC application and analysis of disease processes.

**Table 1.** Potential defects in regulatory T cells in autoimmune diseases [9,10].

Imbalances in peripheral effector and regulatory T cells due to defects in thymic selection Genetic defects inducing failed Treg function or inadequate Treg activity Overwhelming of Treg responses due to epitope spreading in autoimmune diseases, Deficient IL-2 (required for Treg development) Low CD25 expression (hence reduction of IL-2 signalling) Defective conversion of naive T cells to adaptive Tregs (due to IL-10 or TGF-beta deficiency) APC maturation defects leading to altered T cell activation and altered development of tolerogenic phenotype Hyper-costimulation by APCs leading to pathogenic T cells rather than tolerogenic phenotype Aberrant cytokine milieu leading to Treg suppression

The transfer of autologous Tregs to suppress immune responses has already been demonstrated experimentally in SLE and other autoimmune diseases such as diabetes mellitus [11,12]. Regulatory T cells are present at locations of inflammation (e.g., synovial fluid, mucosa) [13] though, if regulatory T cells are obtained from these sites, there may be inadvertent contamination of auto-reactive effector T cells, which could lead to unintended inflammatory consequences from therapeutic reinfusion of collected cells. Once isolated, it is technically challenging to induce these regulatory T cells to proliferate exogenously, which places limits on the application of harvested Tregs from patients for use in therapeutic treatments.

The ability to instead induce functional Tregs rather than needing to collect them, has been demonstrated from iPSCs *in vivo* [14]. These cells produced the immunoregulatory cytokines TGF beta and IL-10, thus producing a population of presumably functional Tregs. In a promising find, both allogeneic and autologous transfers of these iPSC derived Tregs demonstrated clinical efficacy, by reducing disease incidence and clinical severity scores in collagen-induced arthritis (CIA), an inducible mouse model of RA.

#### *3.2. Dendritic Cells*

Dendritic cells are highly proficient APCs that are potent in stimulating naive T cells during the primary immune response [15]. Numerous abnormalities in dendritic cells have been noted in patients with autoimmune diseases, including variations in cells proportions, differences in cytokine receptor expression particularly inhibitory receptors, and increased expression of costimulatory molecules [16,17].

Conventional dendritic cells (cDCs, previously known as myeloid DCs) are extremely efficient APCs, expressing several Toll-like receptors (TLRs) on their surface and producing TNF-alpha, IL-1, IL-6, IL-12, and IL-10 upon stimulation. Under different stimuli, cDCs can demonstrate different tolerogenic phenotypes, inducing antigen-specific unresponsiveness in central and peripheral lymphoid organs, and, therefore, have a crucial role in the induction of immune tolerance [18]. These tolerogenic dendritic cells are characteristically able to induce proliferation of Tregs (which then modulate immune responses to self-antigens), and to induce anergy in auto-reactive effector T cells [18,19]. Depending on the stimuli applied to the cDCs, different tolerogenic phenotypes are demonstrated, with functional differences in the Treg responses that are elicited [10]. Thus, depending on the desired Treg outcome, there is potential to preferentially select these outcomes by altering the particular phenotype of the applied tolerogenic dendritic cell in disease immunotherapy.

For example, Tregs can be induced *in vivo* by NFKB or CD40-deficient DCs. Conventional DCs require the transcription factor RelB to enable priming of the immune system through CD40 and MHC-molecule expression [20,21]. Blocking of RelB and other NFKB family members in cDCs results in induction of Tregs through modified cDC activity, therefore RelB activity is thought to determine the outcomes of antigen-presentation to cDCs. Methods to block RelB activity, and that of other NFKB family members have been developed to produce modified DCs that are consistently tolerogenic through the induction of Tregs [20,22,23]. In murine models of antigen-induced arthritis, modified DCs have been shown to suppress joint inflammation and erosion [24]. As tolerance induction by these DCs has been shown to be dose-dependent and route-independent [22], after induction of inflammatory arthritis by joint injection of methylated bovine serum albumin (mBSA), the mice were able to be subcutaneously injected with modified DCs exposed to mBSA, resulting in a suppression of inflammatory responses in the joints.

Given proof of concept studies using regulatory DCs in immunotherapy have demonstrated a reduction in effector T cell in other autoimmune diseases [25,26] the use of regulatory DCs as autologous immunotherapy is an exciting focus for possible future therapies [10,16,17], particularly in the immunomodulation of the inflammation noted in SLE and RA.

Plasmacytoid dendritic cells (pDCs) constitutively produce anti-viral Type 1 interferons as part of the immune response to viral infections. However, in patients with autoimmune diseases, such as SLE, pDCs are thought to instead make interferons following TLR ligation by endogenously derived nucleic acids [27]. The immune response is, thus, driven not by exogenous infection, but by activity against self-antigens.

Plasmacytoid dendritic cells that produce Type I interferons are found in the tissues of affected organs in SLE and other autoimmune conditions. Type I interferons have activity through several down-stream pathways to increase dendritic cell maturation and activation and, hence, antigen presentation to immune lymphocytes, and non-haematopoietic cell cytokine and MHC expression [6]. This immune activation results in up-regulated inflammation, and a positive-feedback loop with further dendritic cell production of interferon, and resultant anti-self T cell activation and B cell auto-antibody production.

In patients with active SLE, polymorphonuclear lymphocytes (PMNLs) have been shown to up-regulate interferon genes giving an interferon "signature", which correlates with disease severity, and high dose steroids which abrogate this signature induce clinical remission. Depletion of pDCs early in the course of SLE can reduce the clinical and serological evidence for autoimmunity [28]. This evidence indicates that the ability to model the interactions of pDCs would be beneficial to understanding more of the underlying pathogenesis in SLE.

The routine use of dendritic cells for research into the generation of immunomodulation, or for disease modeling *in vitro*, in SLE, RA and other autoimmune diseases is limited by the lack of plentiful and stable dendritic cells of the appropriate phenotype. Peripheral collection of precursors for autologous transfer through plasma exchange is not without morbidity, and the cost and logistics for wide-spread collection may not be feasible. Therefore, while able to be generated from haematopoietic stem cells, regulatory dendritic cells have recently been generated from murine iPSCs [19]. These iPSC-derived regulatory dendritic cells have been shown to have similar morphology to bone marrow derived regulatory DCs, and appeared to have similar activity to bone marrow derived regulatory DCs in not stimulating allogeneic CD4+ T cells, only weakly stimulating allogeneic CD8+ T cells and having similar efficient antigen uptake. What remains is to demonstrate stable phenotype and function, which can then enable comparison of results in clinical trials and other applications to be explored.

Once cells are generated from iPSCs, these need to have a valid functional assessment for tolerogenic properties. Similarly, as there is a theoretical risk for replication of the disease process with autologous transfer of cells, and a demonstrated risk for malignancy with iPSCs, appropriate monitoring and assessments will be required.

#### *3.3. Disease Modelling in SLE or RA*

Theoretically, the potential for disease modelling could be greatly expanded by generating and studying the different tissue lineages from patient-derived iPSCs [3]. While neurological tissue collection remains elusive, methods for expansion of renal specific cells into iPSCs through non-invasive urinary cell collection has been described [29]. Therefore *in vitro* examination of pathological processes using iPSCs derived from affected patients, and, possibly, regeneration of tissue from unaffected patients may both be possible. However, the end-organ damage of SLE is a manifestation of systemic immune dysregulation therefore the targets of therapy or investigation may be more well-focussed on the interactions between cellular populations and an examination of the matrix of effects on tolerance and auto-reactivity. Both SLE and RA are multifactorial in their pathogenesis with a complex interaction between environment and genetics, resulting in the loss of self-tolerance [30,31].

#### **4. Generation of Reparative Tissue in Autoimmunity—Diabetes Mellitus**

Diabetes mellitus is a significant clinical problem with high morbidity and mortality associated with microvascular and macrovascular complications of hyperglycaemia. Arising either from beta cell dysfunction and insulin resistance, or from autoimmune cell-mediated pancreatic islet cell destruction and resultant lack of insulin, treatments are usually aimed at glycaemic control, or reducing insulin resistance. Accurately and consistently replacing insulin at an amount appropriate for associated oral intake can be difficult for patients, with the risk for unstable sugars and hypoglycaemia.

Replacement of pancreatic tissue through tissue donation is in current use, however limited through lack of donors and restrictive through the requirement for life-long immunosuppression. It has been previously pointed out therefore, that treatment for diabetes would ideally renew beta cell function and, hence, insulin for glycaemic control, prevent repeat autoimmune destruction of the new pancreatic tissue, and repair the micro- and macrovascular complications that may have already occurred [32].

The current state of play with iPSCs and diabetes, also detailing concerns of immunogenicity, tumorigenicity, appropriate differentiation, full maturation, stability of function, and successful engraftment have recently been reviewed [33] with much work still required for understanding the basic biology of reprogrammed cells.

However, in terms of current research aspirations, there is great interest in attempting to recapitulate normal pancreatic development and generate pancreatic cell types from pluripotent cells [34]. This would encompass differentiating iPSCs into definitive endoderm, morphogenesis into a three-dimensional structure with contact with appropriate mesenchymal supportive cells to provide required growth and development signals, and then commitment of the pancreatic endoderm to endocrine precursor cells and thence to beta cells that produce the required insulin in a glucose-responsive fashion.

Thereafter, considerations need to be made on prevention of rejection of transplants, potentially preferring patient-specific iPSC generation and autologous transfer [35]. iPSC lines have so far been generated from patients with type 1 and type 2 diabetes, as well as maturity-onset diabetes of the young [36–38].

In terms of functional beta cell production, polyhormonal insulin-expressing cells have been derived from human embryonic stem cells and transplanted for some years now, though whether from insufficient cell volume transfer, or transfer of functionally immature beta cells, while helping fasted blood glucose states, they do not yet consistently ameliorate diabetes in non-fasted mice subjects, or tend to lose insulin-secretion capacity [39–41]. In an alternative line of investigation, when given enough time to develop *in vivo* (90–140 days post transplant), engraftment of pancreatic progenitor cells derived from human embryonic stem cells have been able to secrete insulin, and maintain normoglycaemia in a murine model of induced diabetes up until the grafts are removed [42].

Subsequently, glucose-responsive, insulin-producing cells have been generated from human iPSCs and also shown to have the ability in murine models to reverse hypoglycaemia [43], however, can lose insulin secretion over time [44]. While it is important to remember that there are differences between embryonic stem cells and iPSCs [45], potentially, progenitor pancreatic cells may be developed as well from iPSCs for trials in engraftment, but with the advantages inherent over requiring embryonic cell sources.

#### **5. iPSCs in Autoimmune Neurological Disease—Multiple Sclerosis**

Inducible pluripotent stem cells have been studied extensively in neurodegenerative and neurogenetic disorders, more so currently than for inflammatory neurological conditions, such as multiple sclerosis (MS), however, the final common pathway of neuronal injury and death is better understood in MS than for neurodegenerative conditions. IPSC technology allows potential avenues for therapeutics by regeneration of specific neuronal populations [46] or for exerting an immunomodulatory effect [47], but also allowing more accurate modelling of neurological disease than can be obtained through animal studies [46].

MS is the archetypal and most common disabling autoimmune condition of the central nervous system (CNS), which provides an ideal framework for research and understanding immune dysregulation. MS is a chronic condition, characterised by focal or multifocal inflammatory demyelinating episodes resulting in neurological disability depending on the area of the CNS involved. There are periods of quiescence and recovery in the most common phenotype, known as remitting relapsing MS [48].

The pathogenesis of MS and its triggers are multi-factorial with a complex interaction between genetic predisposition and environmental factors resulting in immune dysregulation. The first risk allele to be identified was the human leukocyte antigen (HLA) class II haplotype HLA-DRB\*1501 in the 1970s [49]. The Genome-wide Association Study (GWAS) has since identified over 50 susceptibility loci [50], many of which encode for pro-inflammatory IL-2 and IL-7 [51], with others encoding for cytokines, such as CXCR5, IL-12A, IL-12ȕ, and IL-12Rȕ1 [48].

The genetic association alone does not explain fully the development of MS with vitamin D3 and Epstein-Barr virus (EBV) both being important environmental factors to consider in MS. Increased latitude is associated with lower serum levels of vitamin D3, due to lower levels of sun exposure, which corresponds with the higher incidence and prevalence of MS in these high latitude countries [48,52] though the effect of vitamin D3 deficiency on adaptive immunity is not yet fully understood. What has also been observed, is that individuals who are seronegative for EBV have almost no risk of developing MS [53], and it has been hypothesised that, through molecular mimicry, EBV may mimic myelin basic protein pathogenic antigens by presentation on HLA-DRB1\*1501, therefore, providing links to both environmental and genetic risk factors [48,54]. Myelin reactive CD4+ T cells secreting interferon gamma are one of many T cell mediators in the pathogenesis of MS [55], with the role of other cell types and cell subsets being also involved, with a reduction in effector function of Tregs in MS patients [56], and a key role of pro-inflammatory T helper 17 (Th17) cells emerging [48,57]. Given the production of oligoclonal bands in CSF, there is a role of B cells in MS pathology, and the understanding of the part played by innate immunity by way of NK (natural killer) cells and dendritic cells in the pathogenesis is evolving [48].

Given the significant effects of MS on affected patients, efforts to provide regenerative or immunomodulatory therapy are highly sought.

Oligodendrocyte precursor cells (OPCs) derived from iPSCs, first described by Onorati *et al*. in 2010 [58], possibly provide an exogenous way in which to remyelinate axons as soon as possible after an episode of acute demyelination, to best protect axons from ongoing inflammation and eventual gliosis. Axonal loss is responsible for the most debilitating functional deficits in the more progressed stages of MS, with this loss followed by retrograde neuronal degeneration [59]. Axonal degeneration not only occurs in chronic lesions, with good evidence now showing axonal injury in acute lesions [60].

Cell replacement with OPCs derived from iPSCs have been shown to be successful in animal studies, with remyelination and amelioration of disability in experimental autoimmune encephalitis (EAE), an animal model of MS [61,62].

Neural precursor cells (NPCs) derived from iPSCs have also been shown in EAE to not only have a regenerative effect, but also an immunomodulatory effect. One study, in which mouse iPSC-derived NPCs were intrathecally transplanted in mice with EAE, exerted a neuroprotective effect, not by differentiating into myelin producing cells, but by producing the specific neurotrophin, leukaemia inhibitory factor (LIF), which supports the *in vivo* survival and differentiation of native oligodendrocytes [63]. LIF has been shown to inhibit the differentiation of Th17 cells through MAP kinase suppression of the cytokine signalling 3 (SOCS3) inhibitory signalling cascade, antagonising the interleukin 6 (IL-6)-mediated phosphorylation of signal transducer and activator of transcription 3 (STAT3) [64], which is essential for the differentiation of Th17 cells, thus limiting CNS inflammation and hence subsequent tissue damage.

Finally, the disease in a dish approach may give unique insights into the study of pathogenesis in neuronal disease and in particular to inflammatory diseases of the CNS, given its inaccessibility. IPSCs have been successfully derived from a MS patient's dermal fibroblasts, and differentiated into astrocytes, oligodendrocytes and neurons with a normal karyotype. The patient-derived neurons showed electrophysiological differences compared with the control cell line, paving the way for a novel approach to the study of MS pathogenesis [65].

#### **6. Conclusions**

Autoimmune diseases are the result of a combination of environmental influences acting on a susceptible genetic background. This causes significant aberrations of self-antigen recognition, lymphocyte activation and differentiation, production of pro-inflammatory cytokines and autoantibodies, and the final end product of tissue and organ damage. Induced pluripotent stem cells technology has the potential to create new safe treatment options, as well as better models to study disease and therapies *in vitro*. Here, we review the so far limited literature in this field. In addition to organ replacement strategies where iPSC technology has been applied, we propose that complex auto-immune diseases require unique immunomodulatory therapy strategies using cellular components and that these components could be made by iPSC technology. Importantly, iPSC technology enables us to produce, differentiate and genetically modify large numbers of immune cells that can be used therapeutically. Prior to the development of such technologies modification of small cell populations with limited *ex vivo* expansion potential was near impossible. Nevertheless, these novel approaches will need to have extensive functional and safety assessments prior to their use in a clinical setting.

Finally, iPSC technology allows for modelling of normal and diseased (based on genetic and epigenetic modifications) cellular growth and development, influences of mutations onto function and clinical phenotype. In the time of personalized medicine iPSC technologies are likely to feature as a key therapeutic tool in auto-immune diseases.

#### **Acknowledgments**

Michael J. Edel (RYC-2010-06512) is supported by the Program Ramon y Cajal, project grant BFU2011-26596 and UWA-CCTRM near miss grant.

#### **Conflicts of Interest**

The authors declare no conflict of interest and support free open access publishing.

#### **References**


protein 1 and its effect on antigen-presenting function in B cells. *J. Virol.* **2002**, *76*, 1914–1921.




#### MDPI AG

Klybeckstrasse 64 4057 Basel, Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 http://www.mdpi.com/

*JCM* Editorial Office E-mail: jcm@mdpi.com http://www.mdpi.com/journal/jcm

MDPI • Basel • Beijing • Wuhan ISBN 978-3-03842-121-4 www.mdpi.com