Induced pluripotent stem cells in the study of neurological diseases

Five years after their initial derivation from mouse somatic cells, induced pluripotent stem (iPS) cells are an important tool for the study of neurological diseases. By offering an unlimited source of patient-specific disease-relevant neuronal and glial cells, iPS cell-based disease models hold enormous promise for identification of disease mechanisms, discovery of molecular targets and development of phenotypic screens for drug discovery. The present review focuses on the recent advancements in modeling neurological disorders, including the demonstration of disease-specific phenotypes in iPS cell-derived neurons generated from patients with spinal muscular atrophy, familial dysautonomia, Rett syndrome, schizophrenia and Parkinson disease. The ability of this approach to detect treatment effects from known therapeutic compounds has also been demonstrated, providing proof of principle for the use of iPS cell-derived cells in drug discovery.

cells by the forced expression of defi ned factors [9][10][11]. Distinct factors, and strategies to induce their expression, have been employed for the generation of iPS cells from a number of human tissues using an array of approaches with varying degrees of effi ciency [12]. To date, however, most patient iPS cell lines have been derived by retroviral transduction of dermal fi broblasts due to their accessibility and relatively high effi ciency of reprogramming.
iPS cells can be coaxed into specifi c cell types by manipu lation of the culture environment. Growth factors, small molecules and extracellular matrix proteins can be applied in a sequential manner to emulate the normal development of the cell lineage of interest. Using this approach, investigators have been able to diff erentiate human pluripotent cells into lineages necessary for model ing neurological diseases, including cholinergic [13,14], glutamatergic [15] and dopaminer gic neurons [16,17], astrocytes [13], oligodendrocytes [18] and Schwann cells [19,20].
Spinal cord cholinergic motor neuron diff erentiation is one of the better studied among the aforementioned cell types and follows the same steps described during normal embryonic development [21]. Th e fi rst step in diff erentiating iPS cells into neurons is inhibition of pathways such as those of transforming growth factor beta and bone morphogenetic protein [22]. iPS cells diff erentiate to neuroepithelia usually within a few days of compound treatment and assume a neural tube-like rosette morpho logy. Th is primitive neuroepithelium can be patterned to ventral spinal progenitors by treatment with retinoic acid and sonic hedgehog or one of its signaling agonists. Retinoic acid is the main signal for Figure 1. Human induced pluripotent stem cells can be diff erentiated into cell types to study neurological disorders. Human induced pluripotent (iPS) stem cells can be diff erentiated into cell types relevant for the study of neurological disorders. Somatic cells from patients with neurological disorders can be reprogrammed into pluripotent stem cells, which in turn can be diff erentiated into distinct neuronal and glial cell types, thus off ering a human cell platform for mechanistic studies and high-throughput screening for diseases of the central and peripheral nervous system.

Phenotype characterization
• ell mor olo r anelle tra ic in • lectro siolo Gene e ression

Mechanistic studies
• arget i enti ication at a s molecules)

Drug discovery
High throughput screening

Cell replacement therapy
neurons to assume a caudal (spinal cord) profi le, while sonic hedgehog deter mines a ventral (motor) identity. Further diff erentiation to mature spinal motor neurons can then be accom plished by addition of specifi c factors, such as brain-derived neurotrophic factor and glial cellderived neurotrophic factor, both of which promote axonal elongation [23]. Th is process usually takes around 3 to 6 weeks depending on the specifi c protocol, and can be monitored using a set of markers including PAX6 (neuro epithelia), OLIG2 (motor neuron progenitors), ISLET1/2 and HB9 (motor neurons), and acetylcholine transferase and synapsin (mature motor neurons), among others. Alternative approaches including the generation of embryoid bodies as an intermediate step have also been described [24]. Consistent with what is seen in normal development, glia cell diff erentiation only occurs after a prolonged time in culture, usually between 6 and 8 weeks.

Modeling neurological diseases using iPS cells
Identifi cation of a disease-relevant phenotypic diff erence between cells derived from patients and from healthy individuals is one of the most challenging aspects of using iPS cells for disease modeling. Th is is particularly relevant for diseases where causative cellular pathophysiology is not clear, such as familial ALS or Alzheimer disease. Even though iPS cells have been derived from patients with a number of neurological diseases (summarized in the next sections and in Table 1), initial work has focused on modeling neurodevelopmental disorders -in particular, those with known genetic causes. Modeling genetically complex, late-onset diseases is probably more challenging, and may require exposing the cells to biological, chemical or environmental stressors to reveal pathological phenotypes. Th e examples discussed below demonstrate the value of iPS cell-based models for identifi cation of disease mechanisms, discovery of molecu lar targets and development of phenotypic screens for drug discovery.

Monogenic early-onset disorders
Spinal muscular atrophy SMA (OMIM: 253300) is an autosomal recessive disease that aff ects one in every 6,000 to 10,000 live births, making it the most common neurogenetic disorder of infancy. SMA is caused by a decrease in levels of survival of motor neuron (SMN) protein due to deletions of the SMN1 gene. Even though SMN protein is ubiquitously expressed, its defi ciency leads to a loss of motor neurons of the spinal cord ventral horns and consequent denervation of axial and limb muscles, represented clinically by muscle atrophy and weakness, dysphagia and respiratory failure in severe cases [25]. Th e clinical phenotype of SMA is modulated by the expression level of SMN2, a paralog almost identical to SMN1. SMN2 generates low levels of the SMN protein that are not suffi cient to prevent loss of motor neurons. Past studies have largely relied on animal models or unaff ected cell types such as patients' fi broblasts, providing limited insight into the disease mechanism and yielding ineff ective drug treatments. In the fi rst proof-ofprinciple study using iPS cells to model a disease, Ebert and colleagues generated iPS cells from a SMA patient and used them to derive motor neurons [14]. Interestingly, the authors found comparable size and number of motor neurons at 4 weeks of diff erentiation between the SMA and control cultures. By week 6, however, the SMA motor neurons were selectively reduced in number and size when compared with the control cells -suggesting that SMA motor neurons developed normally, but were more susceptible to degeneration. Th e authors identifi ed a reduction in SMN aggregates (also termed gems) in SMA motor neurons, consistent with the reduced levels of SMN in these cells. Th e administration of valproic acid and tobramycin led to the increase of gems in SMA iPS cells. While this study did not show whether these compounds can elevate SMN levels or rescue the loss of patient-derived motor neurons, it provided an important validation for the utility of iPS-derived patient cells to model disease.

Familial dysautonomia
Familial dysautonomia (FD) is one of the hereditary sensory and autonomic neuropathies (type III, or Riley-Day syndrome; OMIM: 223900). FD is an autosomal recessive disorder almost exclusive to individuals of Eastern European Jewish origin, aff ecting one in every 3,600 live births in this population. Clinically, it is characterized by feeding diffi culty, alacrimia, orthostatic hypotension without compensatory tachycardia, and decreased pain and temperature perception. FD is usually fatal, with only one-half of the patients reaching adulthood, even with the best standard of care [26].
FD is caused by mutations in the IKBKAP gene [27] that lead to reduced transcriptional elongation of several target genes, some of which are required for cell motility [28]. In a recent study, Lee and colleagues generated iPS cell lines from three patients with FD and demonstrated several disease-relevant features specifi c to the patients' cell lines, including misregulated inhibitor of kappa light polypeptide gene enhancer in B cells, kinase complexassociated protein (IKBKAP) expression, defective neuronal diff erentiation and a decrease in FD neural crest precursor migration [29]. By comparing gene expression profi les of healthy and patient-derived neural crest precursors, genes involved in peripheral neurogenesis and neuronal diff erentiation were found to be diff erentially expressed in FD cells, providing insight into the molecular mechanism(s) of the disease. Using iPS cell-derived neural crest cells as a drug screening platform, a partial rescue of the disease phenotype was achieved after administration of kinetin, a plant hormone previously shown to reduce levels of the mutant IKBKAP splice form in FD-derived lymphoblast cell lines. Kinetin treatment of patient's cells signifi cantly reduced the mutant IKBKAP splice form and increased the number of diff erentiating neurons; however, the level of increased IKBKAP did not lead to rescue of cell motility. Even though the identifi ed compound only partially rescued the disease phenotype in this cellular model of FD, this study demonstrates the value of patient cellbased disease models for drug discovery using phenotypic screens, as well as for identifying novel molecular targets and disease mechanisms.

Rett syndrome
Rett syndrome (OMIM: 312750) is an X-linked autism spectrum disorder characterized by stagnation of developmental skills starting between 6 and 18 months of age, followed by developmental regression, hypotonia, seizures and autistic behavior. Aff ecting one in 10,000 to 20,000 females [30], it is caused by mutations in methyl CpG binding protein 2 (MeCP2), a protein involved in epigenetic and transcriptional regulation of a number of genes [31,32].
In a recent study, Marchetto and colleagues developed iPS cell lines from four female Rett patients, diff erentiated them into neurons and compared them with neurons derived from healthy individuals [15]. While no diff erences were observed in neurogenesis, mature Rett neurons were smaller with fewer dendritic spines and less glutaminergic excitatory synapses. Of note, this phenotype could be modulated by overexpression or knockdown of MeCP2 in neurons derived from control iPS cells, suggesting that MeCP2 is a rate-limiting factor in determining the glutaminergic synapse number in human neurons. Neurons derived from Rett iPS cells also demonstrated reduced frequency of calcium oscillations and spontaneous postsynaptic currents, suggesting a defi ciency in neuronal network connectivity. Similarly to the FD study, the authors identifi ed compounds that partially rescued the disease phenotype in patientderived cells. Th e same group has recently used iPS cells from Rett patients to investigate the role of MeCP2 in modulating long interspersed nuclear elements in neurons, providing yet another example of examining disease mechanisms in patient iPS cell-based models [33].

Late-onset disorders
Amyotrophic lateral sclerosis ALS (or Lou-Gehrig's disease) is the most common motor neuron disease, with a prevalence of one to two per 100,000 worldwide. ALS is characterized by progressive loss of upper (cortical) and lower (spinal cord) motor neurons, with consequent spasticity, hyperrefl exia and progressive weakness and muscle atrophy. It is a fatal disease with a mean overall survival between 3 and 4 years after presentation. Around 10% of cases have a genetic etiology, and animal models have been created based on genes identifi ed in families with ALS. Unfor tunately, no signifi cant drug development has successfully translated from these studies into clinical practice.
In the fi rst work to demonstrate that patient-specifi c iPS cells could be diff erentiated into motor neurons, Dimos and colleagues developed iPS cell lines from two patients with familial ALS caused by a SOD1 point mutation [13]. Of note, the patients were over 80 years old at the time of the study, demonstrating that iPS cells could be successfully generated even from mature skin fi broblasts of the elderly and diff erentiated into spinal motor neurons. Recently, Boulting and colleagues established a test set of 16 iPS cell lines from fi ve healthy controls and two patients with SOD1 familial ALS, and demonstrated that all lines showed comparable effi ciency in the generation of electrically active motor neurons [34]. Th e study found line-to-line phenotypic diff erences between distinct iPS cell lines; however, pair-wise comparisons did not reach statistical signifi cance and con cor dance between lines diff erentiated in two independent laboratories was high, suggesting that the iPS cell platform is reproducible enough to allow for detection of consistent disease-specifi c phenotypes. Although an ALS disease phenotype in patient-derived iPS cells has yet to be demonstrated, iPS cell methodology enables us to create motor neurons from familial and sporadic ALS patients, and to identify common and diverse cellular disease phenotypes in diff erent patients.

Parkinson disease
Parkinson disease (PD) is the second most common neurodegenerative disorder, aff ecting more than 6 million people worldwide [35]. It is characterized by selective loss of dopaminergic neurons in the substancia nigra pars compacta of the midbrain. PD is clinically defi ned by resting tremor, reduced spontaneous movements (bradiky nesia), rigidity and postural instability. A group of nonmotor PD-related symptoms has been increasingly recognized [36], suggesting that other neuronal cell types may also be aff ected. Although PD is a treatable condition, neurodegeneration progresses despite sympto matic control, worsening symptoms and eventually reduc ing therapeutic effi cacy. Dopaminergic neurons, the main cell population aff ected by PD, have been diff eren tiated from patient-derived iPS cells [16,17,37,38]. Th ese neurons were successfully transplanted into rat brains, integrated to the neuronal circuitry, survived in signifi cant numbers 12 weeks after transplantation and improved clinical phenotype as defi ned by a reduction of ampheta mine rotational asymmetry [17,37], closely replicating previous experiments using human embryonic stem cell-derived dopaminergic neurons [39][40][41].
In a recent study, iPS cells were generated from a patient with a homozygous point mutation in the leucinerich repeat kinase-2 (LRRK2) gene, the most common cause of familial PD [38]. Dopaminergic neurons derived from these iPS cells demonstrated increased expression of stress-response genes, including HSPB1, NOX1 and MAOB, increased α-SYNUCLEIN levels and oversensitivity to stress agents, such as peroxide and 6-hydroxydopamine. Seibler and colleagues recently derived iPS cells from patients with familial PD caused by mutations in the PTEN-induced putative kinase (PINK1) gene [42]. PINK1 is an outer mitochondrial membrane protein believed to regulate the translocation of PARKIN, another protein associated with familial PD, into damaged mitochondria. Patient iPS cell-derived dopaminergic neurons exhibited mitochondrial dysfunction that was alleviated by introduction of wildtype PINK1. Taken together, these data suggest that key features of PD pathophysiology could be recapitulated using the iPS cell approach. Potential disease mechanisms identifi ed in dopaminergic neurons derived from iPS cells of familial PD patients could be further studied in cells derived from patients with sporadic forms of PD to help establish common downstream pathways amenable to therapeutic intervention.

Schizophrenia
Schizophrenia is a devastating neuropsychiatric disease characterized by long duration of symptoms, delusions, lack of motivation, reduction in spontaneous speech and social withdrawal, and few aff ective symptoms [43]. Disease onset is usually in adolescence and early adulthood, which causes signifi cant human and fi nancial burden to patients, family and society as a whole [44]. Th e pathophysiology of schizophrenia is complex, including environmental as well as strong genetic components [45]. As with other neuropsychiatric conditions, generation of reliable animal models is limited and problematic [46]. A recent study demonstrated disease-specifi c phenotypes in iPS cell-derived neurons from four patients with schizophrenia, including reduced neurite density, neuronal connectivity and glutamate receptor expression, and altered gene expression of components of the cyclic AMP and WNT signaling pathways [47]. Of note, both neuronal connectivity and gene expression abnormalities were improved after a 3-week treatment with the antipsychotic loxapine.

Other neurological diseases
iPS cells have also been generated from patients with Duchenne and Becker muscular dystrophy [48,49], Huntington disease [48,50], and the genomic imprinting disorders Angelman syndrome and Prader-Willi syndrome [51,52]. Although the resultant iPS cell lines carried the basic genetic abnormality for each disorder, no specifi c phenotype was described under standard culture conditions. However, several fi ndings from these studies are noteworthy. Striatal neurons derived from Huntington disease iPS cells demonstrated enhanced caspase 3/7 activity after growth factor withdrawal [50]. iPS cells derived from patients with Angelman syndrome and Prader-Willi syndrome -neurodevelopmental disorders caused by lack of expression of genes contained in a specifi c region of chromosome 15, and defi ned by the parental origin of the aff ected genetic material (imprinting) -maintained the appropriate DNA methylation imprint following reprogramming [51,52], validating the use of the iPS cell model in the investigation of imprinting diseases.
Recently, Kazuki and colleagues corrected the genetic abnormality in fi broblasts from a patient with Duchenne muscular dystrophy, due to a deletion of exons 4 to 43 of the human dystrophin gene, using a human artifi cial chromosome with a complete genomic dystrophin sequence [49]. At 2.4 megabases, DYSTROPHIN is the longest known gene, making gene replacement therapy particularly challenging, especially for patients with long deletions. Th e authors successfully derived iPS cells from the corrected fi broblasts, demonstrating the potential for combining gene therapy and iPS cell technology to generate patient-specifi c rescued cell lines for eventual use in cell replacement therapy.

Challenges and limitations
Despite the rapid progress in applying iPS cell technology to disease modeling, this promising platform is still in its infancy. Several issues remain to be tackled before iPS cells can be used as reliable models of acquired, multifactorial disorders and, eventually, as treatment strategies in regenerative medicine.
One immediate challenge is in using iPS cells to produce relevant diff erentiated and functional cell types. Current diff erentiation protocols attempt to mimic embry onic specifi cation and patterning; for example, using signaling molecules to dial in the desired rostral/caudal and dorsal/ventral location. Th is approach, how ever, generally results in a heterogeneous cell population. While these mixed populations could be considered co-cultures in which, particularly, neurons are more amenable to long-term maturation and survival, they also present a possible challenge to phenotype identifi cation.
Simple biochemical and gene expression analyses cannot be performed across cultures without careful normalization for cell types and their proportions present, which may limit the study of conditions exclusively or preferentially aff ecting one cell type. However, approaching the diff erentiated culture similarly to a primary explant culture, such as dorsal root ganglia cultures where multiple cell types coexist, may be a useful strategy. In this approach, the heterogeneity of diff erentiated cultures is turned into an advantage where the cell type of interest can be studied within a broader milieu; for example, motor neurons with spinal cord interneurons and glial cells.
Th e use of cell type-specifi c reporter genes allows for identifi cation and characterization of the target cell while preserving functionally meaningful interactions between neuronal and non-neuronal cells. Recently, new techniques to introduce reporter genes into cells have become available, including bacterial artifi cial chromosomes with fl uorescent reporters [53] and zinc fi nger nucleases [54]. Zinc fi nger nuclease technology allows for the effi cient and rapid production of knockin reporter cell lines, wherein sequences encoding fl uorescent reporter proteins can be put under the control of any endogenous regulatory region. Such a labeling approach can in principle allow for any cell type to be identifi ed or isolated, and the insertion of multiple fl uorescent reporters in the same line would potentially allow for cell diff erentiation, maturation and function to be monitored in real time.
Another approach to study the cell type of interest in a complex culture would be to isolate the desired cell type at the end of diff erentiation using techniques such as fl uorescence-activated cell sorting or magnetic bead separation. While combinatorial cell surface markers are well validated for the hematopoietic system, however, identifying surface markers specifi c for the target cell can be challenging, as is the case for spinal cord motor neurons. Which of the aforementioned strategies for analyz ing heterogeneous cultures diff erentiated from iPS cells will prove to be the more adequate to characterize particular disease-relevant phenotypes is a matter for further study.
It remains unclear whether the iPS cell platform will be able to replicate the more complex, multifactorial pathophysiology of late-onset neurodegenerative disorders. It is possible that in these conditions a disease-relevant pheno type would only appear after a long quiescent period, hindering the use of iPS cells in the study of lateonset diseases. Diverse chemical, genetic or environmental stressors could be applied in such instances, however, in order to mature or age cells if necessary to reveal a phenotype. Additionally, some pathophysiology may require at least a partial recapitulation of central nervous system architecture. For example, possible defects in axonal transport in projection neurons might only be recapitu lated in vitro when neurons are allowed to extend axons of signifi cant length and complexity.
Another related issue, inherent to cell culture platforms, is the inability of the iPS cell model to replicate disease mechanism at the tissue or system levelsincluding, for example, protein deposition or infl am mation. On the other hand, the possibility to study a more isolated system may allow investigators to detect the initial steps of a disease process, otherwise superimposed to other subsequent responses. For example, while the iPS platform will probably not be able to replicate the complex anatomical and functional interactions between the distinct cell types aff ected by PD, the recent report of mitochondrial dysfunction in iPS cell-derived dopaminergic neurons from a specifi c familial form of PD demonstrates how this system can detect discrete cellular dysfunc tion that could otherwise be masked by end-stage changes in pathological specimens [42].
In spite of the challenges for harnessing its true potential, iPS cell technology is likely to prove advan tageous for building novel human disease models. Diff erentiation protocols must be further improved while novel culture conditions needed to support iPS cell-derived cells and investigate their phenotypes are developed.

Conclusions
Th e development of iPS cell technology is opening a new avenue for the study of human, disease-specifi c, neuronal and glial cells that promises to revolutionize the neuroscience fi eld. Since the publication of Takahashi and Yamanaka's seminal paper 5 years ago [9], iPS cell lines from more than a dozen distinct neurodevelopmental and neurodegenerative diseases have been established and specifi c disease phenotypes are starting to emerge. Future studies will probably focus on validating these disease phenotypes in platforms that will allow for the screening of therapeutic compounds and the discovery of biologic mechanisms underlying neurological diseases.
Th e widespread availability of human disease-specifi c cells will allow investigators the unprecedented opportunity to conduct mechanistic studies and determine causation in a human model system, rather than just correlation. Th is will allow in vitro phenotypes to be linked to disease pathology, enabling a better understanding of therapeutic manipulations that might lead to a disease-modifying eff ect.
Developing and validating new techniques to reprogram somatic cells into iPS cells without viral integration and to correct genetic abnormalities ex vivo are the next step in the eff ort to apply iPS cell technology in regenerative medicine, and are currently an active area of research. One can envision a near future where iPS cells will be used as a screening tool for personalized medicine and as a reservoir for cell replacement therapy.

Competing interests
JTD and MG are employees of iPierian, Inc. MAS declares that he has no competing interests.

Authors' contributions
The present review was written and edited by MAS, MG and JTD.