Are reprogrammed cells a useful tool for studying dopamine dysfunction in psychotic disorders? A review of the current evidence

Abstract Since 2006, reprogrammed cells have increasingly been used as a biomedical research technique in addition to neuro‐psychiatric methods. These rapidly evolving techniques allow for the generation of neuronal sub‐populations, and have sparked interest not only in monogenetic neuro‐psychiatric diseases, but also in poly‐genetic and poly‐aetiological disorders such as schizophrenia (SCZ) and bipolar disorder (BPD). This review provides a summary of 19 publications on reprogrammed adult somatic cells derived from patients with SCZ, and five publications using this technique in patients with BPD. As both disorders are complex and heterogeneous, there is a plurality of hypotheses to be tested in vitro. In SCZ, data on alterations of dopaminergic transmission in vitro are sparse, despite the great explanatory power of the so‐called DA hypothesis of SCZ. Some findings correspond to perturbations of cell energy metabolism, and observations in reprogrammed cells suggest neuro‐developmental alterations. Some studies also report on the efficacy of medicinal compounds to revert alterations observed in cellular models. However, due to the paucity of replication studies, no comprehensive conclusions can be drawn from studies using reprogrammed cells at the present time. In the future, findings from cell culture methods need to be integrated with clinical, epidemiological, pharmacological and imaging data in order to generate a more comprehensive picture of SCZ and BPD.


Introduction
Reprogrammed cells offer a window into a previously inaccessible sphere of neuro-psychiatric disorders filling a gap between in vivo observations of clinical cases and genetic analysis. So far, research on the pathogenesis of psychiatric disorders on a cellular level has been notoriously difficult due to the lack of proper tissue models for testing hypotheses or medicinal compounds. Tissue sampling has given some insight into differential functioning of cells between healthy subjects and patients, but cannot avoid the doubt inherent to drawing conclusions from one tissue to another. No specimen can be routinely taken from the central nervous system. Animal models have unequivocally been helpful in testing compounds and generating hypotheses on the pathogenesis of various psychiatric disorders. However, as psychiatric disorders are complex and multifactorial, it is unlikely that a single animal model will be able to fully reproduce their pathophysiology (Jones et al., 2011). Moreover, with somatic disease models, we cannot assess whether the laboratory animal does actually display the psychiatric symptoms we are trying to investigate. While affective disorders comprise many symptoms directly linked to behaviour, all of the Schneiderian first-rank symptoms traditionally used to define core schizophrenia (SCZ) belong to the realm of subjective experience (Schneider, 1957;Tandon et al., 2009). We cannot ask a mouse whether it experiences thought broadcasting or hears voices. Sampling of post mortem tissue avoids uncertainty of pathology as it is possible to select patients based on their clinical history. Those analyses are nevertheless confounded by lack of information on earlier episodes, often lifelong medical treatment, post mortem decay, and by death itself, which, among other things, destroys ion gradients and alters tissue functioning and composition.
Since the publication of a robust method for the generation of induced pluripotent stem cells (iPSC) by forced expression of four transcriptional factors (Klf4, Oct4, Sox2 and c-Myc) by Takahashi & Yamanaka, in (2006) the road to in vitro assays for neuronal cells was paved (Takahashi & Yamanaka, 2006;Takahashi et al., 2007). Those pluripotent cells can be differentiated into all three germ layers, so that easily accessible cells such as fibroblasts or hematopoietic cells can be used to yield neuronal progenitor cells and neurons. Inspired by the successful reprogramming of cells with transcriptional factors, Vierbuchen et al. published a method for directly converting mature somatic cells to functional neurons by employing the factors Ascl1, Brn2 and Myt1l in 2010. Those cells showed signs of mature neurons including neuronal membrane potentials and the formation of functional synapses displaying shortterm plasticity (Vierbuchen et al., 2010). A host of reprogramming techniques, either for direct transformation to a target cell, or to a pluripotent state, has been published since.
Today, the notion of altered dopamine (DA) neurotransmission in SCZ is no longer a hypothesis but supported by a wealth of evidence derived from neuroimaging and genetic studies. Results from the two research fields, found with a diverse array of methods, converge on disturbances in pre-synaptic DA transmission as a key element in the pathogenesis of SCZ. Imaging studies show increased DA uptake and storage in the basal ganglia, and an enhanced release of DA into the extracellular space in patients with SCZ. Large-scale genetic studies show an association of SCZ with gene loci coding for DA receptors and proteins involved in synaptic transmission. However, integrating these findings to understanding the impact of genetic and environmental risk factors for SCZ at a cellular level emerges as a one of the key challenges on our way to new drug targets. Research in cultured neurons has the potential to help closing this gap. Instead of just antagonizing DA D 2/3 receptors at the postsynaptic membrane, alternative ways for restoring DA transmission at the pre-synaptic level will hopefully lead to more efficient and better tolerated ways to treat psychotic symptoms in major psychiatric disorders.
In this review, we the summarize findings from cell culture models of SCZ and bipolar disorder (BPD) and attempt to put these findings into the context of current hypotheses. We will first summarize findings from cell culture models suggesting altered synaptic transmission, then we describe evidence focussing on altered cell maturation and early neurodevelopment, and then we report on literature focussing on cellular energy metabolism. An overview on details and limitations of current reprogramming methods is found in the appended 'Techniques of cell reprogramming' section.

Schizophrenia
Schizophrenia is a psychiatric disorder that affects roughly 1% of the population worldwide (Saha et al., 2005). The clinical presentation of the disorder is highly heterogeneous, commonly comprising positive symptoms such as delusions or hallucinations, negative symptoms such as poverty of speech or motivational disturbance, cognitive symptoms such as impaired working memory or executive malfunctioning, and affective symptoms such as depressed or hypomanic mood (van Os & Kapur, 2009). A single patient can display a plurality of symptoms, so that two patients, each being diagnosed with the same disorder, need not share many psychopathological traits.
Twin studies have shown that SCZ has a high heritability of up to 90% (Sullivan et al., 2003), making it a natural target for cellbased assays. There is a limited number of loci of relatively high impact such as Disrupted in schizophrenia 1 (DISC1), first characterized as a risk factor in a Scottish family (Millar et al., 2000), or the deletion 22q11.2, which causes the DiGeorge syndrome (Lindsay et al., 1995). A vast number of low impact risk loci have been identified over time, each contributing only marginally to the manifestation of the disorder in an individual (Duan, 2015). Yet, a significant proportion of heritability found in twin studies remains unaccounted for (Manolio et al., 2009;Eichler et al., 2010). A significant proportion of SCZ-associated risk loci were significantly enriched by enhancers not only in neuronal, but also in peripheral tissue (Psychiatric Genetics Consortium, 2014). Contributions to the development of SCZ by immunological effectors are supported by the recent observation of the involvement of complement component 4 (C4) in synaptic pruning (Sekar et al., 2016). Significant enrichment was also found in skeletal muscle and pancreatic islands, which implies some involvement of non-neuronal tissues and is in line with findings of disturbed metabolism in SCZ (Psychiatric Genetics Consortium, 2014).
In 2011, Chiang et al. were the first to transform cells derived from patients suffering from SCZ to pluripotent stem cells utilizing an integration-free plasmid vector (Chiang et al., 2011). Since this pioneering work, a host of studies has dealt with patient-derived reprogrammed cells. Most frequently, skin fibroblasts were harvested by punch biopsy and consequently transformed into iPSCs and differentiated into neuronal progenitor cells (NPCs) and a neuron-like state. Robicsek et al. (2013) were, to the best of our knowledge, the only ones to prefer hair follicle keratinocytes over fibroblasts for their more efficient reprogramming, availability and ectodermal identity. Apart from low numbers of independent donors, a tendency to recruit patients displaying strong signs of pronounced genetic background such as high-risk mutations, high-incidence families and early childhood onset of the disorder can be noted. Meir Jacobs' review from 2015 gives an overview about possibilities and limitations of iPSC-based studies in the field of SCZ (Jacobs, 2015). A technical review is provided by Duan, (2015).

Bipolar disorder
Bipolar disorder (BPD) is characterized by recurrent episodes of depression and mania. Mania in particular is often accompanied by psychotic states. Similar to SCZ, there is a high genetic contribution to the disorder. In twin studies, heritability between 59 and 87% was noted (Smoller & Finn, 2003). Among the available treatment modalities, Lithium (Li+) displays the greatest efficacy and a robust anti-suicidal effect (Cipriani et al., 2013). Li+ interacts with cyclic adenosine mono-phosphate (cAMP) response element-binding (CREB) protein activity, the membrane potential of neurons via the Na+/K+ ATPase, monoamine neurotransmitters, second messenger cascades, glycogen synthase kinase (GSK) activity and transcriptional factors (Alda, 2015). In the publications concerning reprogrammed cell models derived from BPD patients reviewed here, response to Li+ responsiveness plays a prominent role.
For an overview on patient-derived reprogrammed cell studies in the field of SCZ and BPD, see Tables 1 and 2.

Schizophrenia
Alterations of monoamine neurotransmitter systems, most noteworthy DA, play a key role in the pathophysiology of SCZ (Howes et al., 2015) and might be a suitable target for cell-based assays. The DA hypothesis of SCZ was coined over 40 years, ago when the DA depleting properties of reserpine (Carlsson et al., 1957) and the DA receptor blocking properties of antipsychotic drugs were linked to improvement of positive symptoms in patients with SCZ (Van Rossum, 1966). Over the years, evidence in favour of disturbed DA transmission as a key feature of the disorder has accumulated. Neuroimaging studies with positron emission tomography (PET) and  single-photon emission computed tomography (SPECT) have identified the pre-synapse as the most probable origin of DA dysregulation (Howes et al., 2012). The cultivation of dopaminergic midbrain neurons thus holds great promise for modelling key pathophysiological traits of SCZ. There are some findings regarding alterations of the dopaminergic system derived from non-reprogrammed cells. Utilizing PC12 cells, a cell line derived from rodent adrenal tissue, Kantor et al. (2002) succeeded in inducing amphetamine sensitization, a phenomenon endogenously occurring in patients with SCZ (Howes et al., 2012). The same group showed protein kinase C (PKC) and mitogen-activated protein kinase (MAP kinase) to be necessary for amphetamine sensitization to occur (Park et al., , 2003 The observation of a biological event thought to be of relevance for DA dysfunction in SCZ in a non-human, non-neuronal cell line supports the hope for cell-based assays to further elucidate alterations of the DA system in SCZ. Nevertheless, impairments of DA transmission could not yet be reliably modelled in reprogrammed neurons derived from patients with SCZ. Hook et al. (2014) described a more than 2-fold elevation in the number of tyrosine hydroxylase (TH) positive (TH+) neurons in cells derived from patients with SCZ compared to those derived from healthy subjects. Moreover, they observed elevated K+-induced release of DA, epinephrine and norepinephrine in patient-derived cells. Robicsek et al. (2013) observed severely hampered differentiation to dopaminergic and glutamatergic neurons in SCZ-derived cells. None of the cells expressed the dopamine transporter (DAT). Tyrosine hydroxylase and b3-tubulin were expressed only at minuscule levels. Dopaminergic neuronal cells displayed SCZ NPCs display reduced miR-9 levels. A subset (~50%) of SCZ are below 25% quantile of healthy subjects' NPCs. Perturbed miRNA-9 levels can be observed in NPCs and early neurons but not after prolonged (6 weeks) maturation miRNA-9 is correlated with impaired radial migration of NPCs in SCZ. Overexpression of miRNA-9 rescues migration phenotype. Non-genomic as well as genomic causes might affect altered miRNA-9 Altered miRNA-9 expression affects multiple genes on transcript and protein level   Deficits in depolarization induced vesicle release and impaired synaptic transmission were reported by Wen et al. (2014) The investigated cell cultures comprised more than 90% cells expressing vGLUT1 or alpha Ca++/calmodulin-dependent protein kinase II (a-CAMKII) neurons derived from two patients suffering from SCZ and major depression from a pedigree carrying a frame shift mutation in DISC1. a-CAMKII has been found to be of relevance for the induction of sensitization to amphetamine (Steinkellner et al., 2014). A causative relationship between DISC1 mutation and synaptic dysfunction was established by generating knock-in iPSC lines of an unaffected family member and a non-related control .  examined a set of five SCZ cases, four of which originated from high-incidence families. Alongside with the reduced expression of the post-synaptic marker PSD-95, which was later found to be attenuated also by Robicsek et al. (2013), reduced neuronal connectivity was observed employing a rabies virus assay. Alterations in the expression of genes relevant to glutamate, Wnt signalling and cAMP signalling, were noted by the authors. Reduced synaptic connectivity was rescued by treatment of induced neurons with the antipsychotic loxapine. Loxapine also increased the expression of neuregulin 1 (NRG1) and various glutamate receptors (GRIK1, GRM7 and GRIN2A), as well as the expression of ADCY8, PRKCA, WNT7A and TCF4. No positive effects on connectivity were observed with clozapine, olanzapine, risperidone and thioridazine .

Bipolar disorder
In a prospective design, Mertens et al. (2015) observed six patients with BPD on Li+ monotherapy. Notably, reduced cellular activity as described by Yu et al. was not detected here, but rather an increase in excitability, although identical methods were employed. This indicates a disease-specific alteration. Cellular hyper-excitability in BPD-derived cells was attenuated by Li+ treatment in clinical Li+ responders, but not in Li+ non-responders. The latter displayed attenuation of electrophysiological alterations by lamotrigine. The authors noted much larger changes in gene-transcription in Li+ responders when compared to Li+ non-responders bringing gene expression closer to what is found in healthy subjects. Among the rescued genes, there were 84 directly involved in putative pathological processes, among those PKA/C regulation, neuronal firing rate and Ca++ signalling (Mertens et al., 2015).
Li+ was shown to reduce Ca++ transient and wave amplitude significantly more in BPD-derived neurons than healthy subjectsderived neurons at 12 weeks post-differentiation (Chen et al., 2014).

Findings suggesting disturbed neurodevelopment
As iPSC models mimic neuronal development from a very early stage, perturbations of normal developmental steps can be observed on a single-cell and cell cluster basis.
Schizophrenia SCZ is sometimes conceptualized as a neurodevelopmental disorder most commonly manifesting in late adolescence and early adulthood (Keshaban et al., 1994;Laurelle, 2000). Gene-ontology analyses of genes differentially expressed in SCZ and healthy subjects revealed alterations in pathways involved in 'neurogenesis', 'neuronal differentiation' (Pedrosa et al., 2011) and 'brain development' Brennand et al., 2015). Two authors describe delayed transformation and maturation of SCZ-derived cells (Pedrosa et al., 2011;Yu et al., 2014). Murai et al. (2016) found increased micro-RNA-219 expression which led to reduced neuronal stem cell proliferation in SCZ patients carrying the DISC1 mutation. An examination of microRNA expression in patients with SCZ carrying the del(22q.11.2) mutation was conducted by Zhao et al. (2015). Thirty-two microRNAs were significantly upregulated, while 13 were downregulated. Gene-ontology analysis revealed, among other findings, targets in various clusters important to neurological and psychiatric disease. Yoon et al. (2014) observed that single nucleotide polymorphisms (SNPs) in mediators of WAVE signalling (ACTR2/Arp2) interact epistatically with CYFIP1 deletion (a gene involved in del(15q11.2)) to increase the risk for SCZ. This epistasis could be mediated by disturbances in apical polarity and adherence junctions in affected neurons. Other authors have observed alterations in gene transcription related to temporo-spatial cell identity and adhesion Pedrosa et al., 2011).
Altered expression of miR-34 family expression was noted in induced neurons derived from SCZ and schizoaffective (SZA) patients carrying the del(22q11.2) mutation without reaching genome wide significance levels (Zhao et al., 2015). Passeri et al. (2015) observed a significantly higher transformation efficiency in cells derived from patients carrying the del(16p.11.2) mutation compared to healthy subjects using single-step transformation in an improved medium.
Abnormalities in cell morphology have been described in cells derived from SCZ patients, however, no clear SCZ-related phenotype could yet be established. Two studies noted reduced neurite outgrowth Robicsek et al., 2013). Neurons derived from DISC1 carriers displayed increased soma size and neurite length 2 weeks after transformation but not 4 weeks after transformation . Cells differentially expressed genes from the gene-ontology categories 'brain development', 'synaptic transmission' and 'dentritic spine'. Topol et al. (2015), noticed over-expression of translation-related proteins in SCZ-derived NPCs, but could not detect any abnormalities in cell size that could account for this finding.
Cell motility plays a key role in brain development so that altered cell migration could have neurodevelopmental implications in SCZ. Fan et al. (2013) describe markedly increased mobility and reduced adhesion in SCZ-derived olfactory neurospheres. Furthermore, SCZcells proliferated more readily. Hypermobility was rescued by focal adhesion kinase inhibition. Contradictory to these findings, Brennand et al. (2015) observed reduced mobility in human-induced neuronal SCZ-cells. In an earlier observation by Pedrosa et al. (2011), gene ontology of significantly differently expressed genes in induced SCZneurons revealed 'cell adhesion' as an altered functional cluster. In NPCs derived from a patient with SZA, impaired migration, which coincided with preferential expression of a deficient allele of CNTNAP2 was observed. The patient's not affected father, who carried the same genetic deletion did not display aberrant cell migration while primarily expressing the wild-type allele (Lee et al., 2015). Reduction of NPC migration in SCZ is further corroborated by data from two relatively large cohorts of Topol et al. An overall decreased expression of microRNA-9 in SCZ was driven by a subset of patients below the 25% quantil of healthy subjects. miRNA-9 was correlated with neurosphere migration, and the phenotype of impaired migration in SCZ was attenuated by over-expression of miRNA-9. Marked differences between SCZ and healthy subjects in miRNA-9 expression could be observed in NPCs and early neurons, but not at 6 weeks after transformation. miRNA-over-expression in healthy subjects as well as SCZ-derived NPCs resulted in a trend towards alterations in the protein expression in pathways related to 'actin cytoskeleton', 'protein localization' and 'RNA processing' (Topol et al., 2016).

Bipolar disorder
A variety of regulatory and neurodevelopmental alterations are thought to play a role in BPD. The review of O'shea & McInnis (2016) gives an overview on neurodevelopmental alterations in BPD with a focus on iPSC models. Madison et al. noted significant reduction of cell proliferation in the bipolar offspring of two healthy parents. In four of six iPSC lines derived from BPD, no differentiation to NPCs was achieved. Propagation of NPCs past four or five passages was laborious and could only be achieved for a subset of iPSC lines of BPD. Proliferation of BPD-derived NPCs was reduced, and paired box protein Pax-6 (PAX6) expression decreased. Gene expression did not significantly differ between BPD and healthy subjects derived fibroblasts and iPSCs, however, in NPCs, 18 genes showed > 1.5 fold difference in expression. Multiple genes involved in spatial identity regulation were altered (NKX6-1, LEF1, NEUROG1, NRG3 and SPARCL1; Madison et al., 2015). Impairment of ventral-dorsal patterning and spatial identity was described by Chen et al. (2014) who observed that cells derived from BPD patients display increased DICER1, NKX2-1 and PAX2 expression (all of which promote ventral cell fate) and reduced expression of EMX2 and HOXA1,6 and 7 (which promote dorsal cell fate). Li+ reportedly up-regulates Wnt signalling (Valvezan & Klein, 2012), thus dorsalizing early NPCs . In cerebellum slices of Li+-free BPD patients, significantly elevated levels of miR-34a were detected. In BPD iPSC-derived neuronal cells, inhibitory effects of miR-34a on the expression of ANK3 and CACNB3 were described together with derogation of pre-and post-synaptic development such as reduced dentritic branching (Bavamian et al., 2015). In non-neuronal cells, miR-34a was earlier described as an inhibitor of Wnt signalling (Hashimi et al., 2009).
Using a high-throughput label-free imaging platform (BIND scanner) in BPD, Wang et al. (2014) could detect significantly higher peak wave length in cells derived from Li+ responders when compared to Li+ non-responders. This suggests greater adhesion of cells derived from Li+ responders when compared to non-responders. None of the other parameters (cell count, cell fraction, and cell perimeter) were significantly different between healthy subjects and BPD patients and Li+ responders and non-responders.

Schizophrenia
Alterations of mitochondrial function metabolic stress are thought to contribute to the development of SCZ (Rajasekaran et al., 2015). Irrespective of treatment history, patients with SCZ differed from healthy subjects in a micro-array analysis of genes relevant to energy metabolism and oxidative stress. In a proteomics-analysis, about half of the significantly altered proteins were related to energy metabolism and oxidative stress (Prabakaran et al., 2004). Employing non-reprogrammed SCZ-derived fibroblasts, Fourier et al.
observed altered cell metabolism in response to oxidative stress in a collective of first-episode SCZ patients. Significant changes were observed in arginine-, extracellular matrix-and collagen-related pathways. Fatty acid metabolism and neurotransmitter-related metabolism were unaltered (Fournier et al., 2014). Alterations in energy metabolism and reactive oxygen species (ROS) have been observed on a genetic, morphological and functional level in human-induced neuronal cell models of SCZ. Paulsen et al. did not note any alterations in ATP-coupled respiration, but rather increased extra-mitochondrial O 2 consumption and elevated levels of reactive oxygen species (ROS). These changes occurred in SCZ-derived NPCs but not in iPSCs or fibroblasts. ROS-levels were rescued by treatment with valproate (Paulsen Bda et al., 2012). In another study, the same group observed elevated levels of potassium and zinc levels in SCZ-derived NPCs again rescued by valproic acid treatment (Paulsen Bda et al., 2014). Increased levels of ROS and concomitantly decreased mitochondrial membrane potentials and DNA damage suggesting increased apoptosis were measured by Brennand et al. (2015) In the same sample, cell-to-cell variability but not overall expression of 70 kd heat-shock protein (HSP70) expression after oxidative stress was significantly elevated in SCZ. This finding is in line with increased variability, but no changes in mean expression of HSP70 expression following ethanol and methyl mercury stimuli in SCZ-derived NPCs observed by Hashimoto-Torii et al. (2014). In experiments performed by Robiscek et al., keratinocytes derived from healthy subjects displayed baseline differences in O 2 consumption as compared to keratinocytes derived from patients with SCZ, while their iPSCs did not differ in this regard. Inhibition of mitochondrial complex-one driven respiration by DA was significantly elevated in keratinocytes, as well as iPSCs and NPCs, derived from SCZ patients. Alterations of mitochondrial network connectivity and distribution of mitochondria were described in SCZ patients (Robicsek et al., 2013). Batalla et al. correlated neuronal glutamate-glutamine (GLX) and N-acetylaspartate (NAA) with apoptotic markers in non-reprogrammed fibroblasts derived from first-episode SCZ patients in a combined cell culture and proton magnetic resonance spectroscopy approach. The authors could establish a relationship between decreased GLX and NAA levels in the anterior cingulate and condensed chromatine in staurosporine-treated fibroblasts suggestive of excitotoxicity in the brain (Batalla et al., 2015).
In an approach integrating induced cell models and functional magnetic resonance imaging (fMRI), D'Aiuto et al. (2015) investigated possible relationships between latent Herpes simplex virus 1 infection and cognitive parameters in healthy subjects and SCZ using induced neurons as a model for possible contributions of infectious diseases to the pathogenesis of SCZ. However, induced neurons and fMRI data were not derived from the same patients in this study.

Bipolar disorder
BPD-derived neurons displayed a peculiar phenotype of overall smaller mitochondria that showed greater function in the JC-1mitochondrial membrane potential assay when compared to healthy subjects. Li+ treatment rescued mitochondrial size but did not significantly alter membrane potential (Mertens et al., 2015).

Conclusions
Most studies investigating patients with SCZ were aiming to observe an aetiologically homogeneous population, focusing on high-incidence families and patients with high impact gene variants, which is a sensible approach for studying subtle alterations of genetic perturbations, possibly giving cues for general mechanisms underlying this complex polygenetic and poly-aetiologic disorder (Marchetto & Gage, 2014). Identification of alterations in well-defined cases may facilitate the examination of larger, more heterogenous populations.
The examination of iPSC-derived DA neurons from a pair of monozygotic twins discordant for Parkinson's disease, while carrying a distinct predisposing mutation, yielded significantly different findings in monoamine oxidase B (MAO-B) activity and DA availability (Woodard et al., 2014). This can be interpreted as proof of concept, suggesting that also non-genetic effects may be observed in reprogrammed cells. The method may thus allow for the identification of biological correlates of environmental contributions to complex disorders, in particular in phenotype-discordant monozygotic twins. As the majority of clinical cases are sporadic rather than monogenetic or familial, many aspects may, however, be overlooked when restricting research to twins discordant for a certain disorders. Attempts to recruit patients with distinct clinical parameters have been made. In particular, observing patients with distinct reaction patterns to medication, such as clozapine resistance (Paulsen Bda et al., 2012) or Li+ responsiveness (Chen et al., 2014;Wang et al., 2014) might be interesting.
In any case, for reprogrammed cells to hold their promise of modelling 'brain disease in a dish' (Marchetto & Gage, 2012) and start playing a prominent role in testing and development of medicinal compounds or pathogenetic studies, more work needs to be done. There are some findings discussed earlier, where loxapine ), valproic acid (Paulsen Bda et al., 2012, 2014, or Li+ (Mertens et al., 2015) could revert alterations observed in cells derived from patients with SCZ or BPD. For the time being, it is unclear whether any of these observations would be predictive of clinical efficacy of a compound, or if the observed alterations translate to pathologies in vivo. There are still plenty of pieces missing in the puzzle of modelling psychiatric disorders in vitro. To be able to interpret findings from reprogrammed cell assays, a robust link between clinical traits and observations in cell culture systems has to be established. In order to achieve comparability between studies, meticulous description of the reprogrammed cells' phenotype and expression patterns is paramount. As elucidated by the example of discrepant findings between the reprogramming of adult cells from patients with SCZ to dopaminergic neurons (Robicsek et al., 2013;Hook et al., 2014) mentioned earlier, a finding or lack thereof might not be interpretable without thorough characterization of the cell. As neither TH nor DAT are truly unique to dopaminergic midbrain neurons (Ciliax et al., 1999), a whole battery of tests is necessary to pinpoint cellular identity. Electrophysiological behaviour, tempo-spatial identity, gene expression, characteristic proteins, pharmacological sensitivity and morphology should, where appropriate, be used to confirm comparability of findings achieved with different reprogramming techniques. Reprogramming remains a volatile procedure, where small alterations in the protocol might change the outcome. In the case of single-step reprogramming to DA neurons, it has been shown that dish surface characteristics influence reprogramming efficiency and cell morphology (Yoo et al., 2015). While it may be impossible to ascertain in how far a reprogrammed cell represents an in vivo counterpart, thorough characterization of its characteristics will help to increase faith in the validity of the model (Srikanth & Young-Pearse, 2014;Hartley et al., 2015).
Data from cell cultures and genetic analysis will need to be integrated with imaging studies, specifically PET and fMRI, to create a more comprehensive picture of psychiatric disorders (Brennand & Gage, 2011). To the best of our knowledge, only two publications combine imaging data and cell culture findings trying to crosslink information in a meaningful way. The study of D'Aiuto et al. (2015) primarily employed cell culture techniques to prove the concept of latent HSV-1 infection in human neuronal cells. Batalla et al. (2015) did not utilize reprogrammed cells, but attempted to directly correlate in vitro with in vivo data. Surprisingly, strong correlations between brain metabolites and behaviour of cultured cells were found. Establishing correlations between single-cell phenomena, clinical features and imaging data in individual patients may facilitate the interpretation of in vitro findings in the future. Many features of cultured cells can be altered by the reprogramming process or remnants thereof, as well as by non-physiological circumstances such as the presence of immature cells in the culture, lack of synaptic contacts and glia, or loss or disturbance of the epigenetic imprint. Furthermore, cell culture and imaging techniques yield profoundly different data. For instance, increased cell size in vitro does not necessarily correspond to altered volumes in MRI studies. When going beyond describing differences between healthy subjects and patients, clear hypotheses regarding pathophysiological phenomena likely to occur on a singlecell basis may be helpful. PET in particular could establish a link between in vivo and in vitro observations by correlating factors such as receptor availability or neuro-transmitter release with genetic and cell culture data. Unlike fMRI, which is limited to observations on a network level, PET is able to visualize targets on a molecular scale. Once cellular and clinical phenotypes have been put into relation, it may be possible to predict clinical outcomes, such as response to medication, based on cell culture models.
If only a specific phenomenon or marker is of interest rather than the process of development and maturation per se, it would be desirable to identify just the right cell type suited for the assay, and to convert cells in a single-step method. This way, the sample size can be increased allowing for the recruitment of a more realistic study population without sacrificing too much statistical power. However, the role of cell maturity and possible effects of the source cells' epigenetic imprint will have to be considered.
A recent meta-analysis of PET imaging data on the DA system in patients with SCZ has yielded large effect sizes for results indicating a pre-synaptic dysfunction of the DA system in SCZ (Howes et al., 2012). There are well-established imaging procedures for quantifying DA release and synthesis in vivo, as well as reliable reprogramming techniques yielding functional dopaminergic midbrain neurons of high purity. The pre-synaptic disturbance of the dopaminergic system therefore seems to be an accessible and promising target for further experiments. As antipsychotic drugs commonly interfere with dopaminergic transmission by blocking post-synaptic D 2/3 receptors, establishing an in vitro model of pre-synaptic DA dysfunction in SCZ might also facilitate drug development or pre-clinical testing of antipsychotic compounds in the future.
On the other hand, many research questions regarding early brain development in SCZ, the development of psychosis, and brain alterations in subjects who are at risk mental state for psychosis remain unanswered. In the future, more complex traits, in particular neurodevelopmental perturbations, could be successfully modelled in organoids (Marchetto & Gage, 2014). Brain organoids comprise multiple neuronal and glial cell types, resembling the brain in early pregnancy (Lancaster et al., 2013;Lancaster & Knoblich, 2014). As of today, development of brain organoids beyond an early stage is hindered by the lack of vascularization, so that later developmental stages remain inaccessible. Some events predisposing to the outbreak of SCZ, such as intra-uterine infection, birth trauma or seasonality of birth occur during pregnancy and might to some extent be modelled in vitro. While some proportion of the genetic loci increasing the risk for SCZ are assumed to impact early brain development, others might not contribute to the appearance of the phenotype until much later in life.

Techniques of cell reprogramming
The process of reprogramming an adult somatic cell is dependent on multiple factors each potentially influencing the induced cells' phenotype and function. Until now, multiple cell types of different origin have been successfully reprogrammed. However, as of yet, no cell type has been shown to be clearly superior. Even while comparatively difficult to harvest, skin fibroblasts remain the most common source, possibly because there are many protocols readily available in the literature. Despite their individual advantages, alternative, cell types such as CD34+ blood cells, squamous cells derived from urine, or hair follicle keratinocytes are as of yet rarely used. The properties, advantages and disadvantages of various cell types are reviewed by (Raab et al., 2014). The impact of the original cells is not confined to traits such as reprogramming efficiency, availability or number of published methods, but stretches into the realm of epigenetics. DNA and histone methylation patterns remain partly that of the tissue of origin, so that iPSCs differentiate most readily to closely related cells (Kim et al., 2011;Vaskova et al., 2013). One recent examination did, however, come to the conclusion that genetic and epigenetic constitution of the donor outweigh cell-type effects (Kytt€ al€ a et al., 2016). By treatment with chromatin-modifying compounds or secondary and tertiary reprogramming, the differentiation-inhibiting properties of epigenetic memory can be attenuated (Kim et al., 2010). However, as a cell's epigenetic imprint might influence a disease-relevant phenotype, such procedures remain controversial.
In order to successfully reprogram adult somatic cells, either to a different cell type or stem cells, the selected transcription factors must be delivered to the cell. Multiple vectors have been reported so far, each bearing benefits and disadvantages, so that it remains an individual choice which system to employ for a specific problem. For a comprehensive analysis of reprogramming vectors, see (Hu, 2014). In addition to high reprogramming efficiency, the ability to enter the desired cell, safety, price and the dynamics of transgene silencing are key points of interest for vectors in research. As there are at the time being no ambitions for the development of cell therapies based on reprogrammed cells in psychiatry, the issues of oncogenic transcriptional factors as well as immunogenicity of many vectors and, to a lesser extent, the mutagenicity of integrating vectors are of limited concern.
In most studies considered for this review, patient-derived cells were first reprogrammed to iPSCs and later differentiated to NPCs and/or any neuronal population. Today, there are various published methods for the generation of patient-derived neurons either through reprogramming to iPSCs, or through direct transformation of adult cells to functional neuronal cells. (Tran et al., 2013) Depending on the research issue, either method has its merits. While iPSCs offer the possibility to observe developmental steps as they would occur in utero, single-step reprogramming techniques have the advantage of being less costly and laborious. Although directly converted cells reach a mature phenotype much faster than iPSC-derived cells (Caiazzo et al., 2011;Kriks et al., 2011), it remains unresolved whether this is a benefit allowing for rapid observation of functional differences, or whether it is a limitation, possibly omitting pathological steps in development while keeping much of the epigenetic imprint of the tissue of origin.
Today, a multitude of reprogramming techniques exists. First, direct conversion of fibroblasts to dopaminergic midbrain neurons by forced expression of the factors published by Vierbuchen et al. (2010) (Ascl1, Brn2 and Myt1 l) together with Lmx1a and FoxA2 (Pfisterer et al., 2011), or solely Mash1, Nurr1 and Lmx1a (Caiazzo et al., 2011) was reported. As of yet, a variety of protocols for the generation of specific neuronal populations have been published. Second, starting from protocols promoting neural induction and functional floor plate formation using human embryonic stem (hES) cells (Chambers et al., 2009;Fasano et al., 2010). Kriks et al. (2011) developed a method for generating dopaminergic neurons with marker profile and distinct electrophysiological features reminiscent of those peculiar of human midbrain development. In this case, the required co-expression of FoxA2 and LMX1 typical of mature dopaminergic neurons is obtained by combining inhibitors of the transcription factors SMAD, sonic hedgehog agonists and GSK3B inhibition-mediated Wnt signalling activation. Along with the optimization of methods to produce 2D cultures representative of specific types of neurons, it is worth mentioning the recent efforts made to generate 3D neuronal model systems that have proven useful to model neuro-developmental diseases (Lancaster et al., 2013;Mariani et al., 2015;Qian et al., 2016). In particular, developing forebrain cerebral organoids showing significant overlap between differentially expressed genes and SCZ risk genes, as well as midbrain cerebral organoids specifically expressing dopaminergic neuron markers were recently obtained from iPSCs (Qian et al., 2016). Although comprehensive physiological characterization is not yet available, midbrain organoids expressing dopaminergic neuron markers are generated starting from previously successful protocols used to model Parkinson's disease (Kriks et al., 2011) and show a relative early appearance of mature TH-expressing neurons. Considering the cell mass-fold expansion typical of 3D organoids, such a protocol would have the advantage of yielding high numbers of neurons in a cost-effective way and relatively short time, representing a valid alternative to 2D cultures. However, a limitation of the system might be the poorly defined identity of the cellular population obtained by this method: At day 65 only about half of the FoxA2expressing neurons were also showing TH expression. While obviously not suitable for certain applications, like transplantation in animal models, the self-organization observed in such organoids might reflect a more reliable approximation of the early steps in the maturation of the developing midbrain and thus represent an advantage in in vitro studies, e.g. when the role of mutations on proper neuronal maturation is hypothesized.
Cell maturity is likely to play an important role in the interpretation of data acquired from induced human neuronal cells. While cells derived from older individuals are reprogrammed at a lower efficiency than cells derived from younger individuals, reprogramming to an iPSC state seems to rejuvenate the cell (Li et al., 2009;Mahmoudi & Brunet, 2012). The relevance of cell donor age is however questioned, as other groups reported efficient reprogramming of fibroblasts from elderly individuals (Marchetto et al., 2010). Furthermore, an increase in iPSCs reprogramming efficiency appears to be achievable from both, high fibroblast passage number and aged donors, respectively, through inhibition of the upregulated cell cycle factor p21, potentially expanding sample typology (Trokovic et al., 2015).
Another point worth discussing are limitations of modelling SCZ through reprogrammed neurons: a population of fibroblast-derived early neurons at 6 weeks differentiation and neuronal progenitor cells derived from patients with SCZ and healthy subjects resembled first trimester neuronal cells regarding gene-expression patterns . The onset of psychotic illnesses such as SCZ and BPD is most common in late adolescence and early adulthood, however, a proportion of patients will develop first symptoms in infancy or much later in life (Kessler et al., 2005;Jones, 2013). Even though SCZ is associated with subtle unspecific deficits in social functioning, motor skills and cognition before a prodromal phase occurs (Murray et al., 2006), and even though some traits like anxiety, attention-deficit and subclinical changes in mood can predate BPD by many years (Serra et al., 2015), those alterations do not occur in all individuals and are not necessarily causally linked to the development of psychosis. Immature cells resembling cell populations found in the prenatal age can most likely be employed to model alterations of cellular functioning very early in the development of those disorders, as discussed above. However, they do not necessarily reflect pathological mechanisms occurring later in life. As an example, considering the role attributed to age as a central aetiological factor in the field of neurodegenerative diseases, progerin treatment has been shown to substantially accelerate the process of maturation and degeneration in iPSCs, resulting in a good model for Parkinson's disease (Miller et al., 2013). Having this in mind, additional cellular model systems should be explored to mimic the effects of age-related effects on the onset of psychotic disorders.