How to make a hippocampal dentate gyrus granule neuron Free

The hippocampal dentate gyrus (DG) and the olfactory bulb are two areas of the mammalian brain where neurogenesis continues to occur throughout life (Eriksson et al., 1998). A recent study using radioactive carbon dating reported a striking annual turnover rate of 1.75% in the human DG where approximately 700 new neurons are added every day (Spalding et al., 2013). These new neurons are generated at the subgranular zone (SGZ) of the DG and integrate into the existing hippocampal circuitry, where they play a fundamental role in learning and spatial memory formation by performing pattern separation on inputs from the entorhinal cortex, an area of the brain located in the medial temporal lobe (McHugh et al., 2007; Zhao et al., 2008; Aimone et al., 2011). Granule neuron cells are the most prevalent type of excitatory glutamatergic neuron in the DG, and their ongoing production in adults is regulated by physical exercise, dietary restriction and enriched environment through upregulation of brain-derived neurotrophic factor (BDNF) (van Praag et al., 1999; Rossi et al., 2006; Zhao et al., 2008).

Aberrant hippocampal neurogenesis is implicated in a number of neurological pathologies, and interventions aimed at regulating neurogenesis have been proposed as potential therapeutic strategies (Jun et al., 2012). Epileptic seizures have been shown to affect neurogenesis and the emergence of abnormal hilar basal dendrites in newborn granule neurons. Normal granule neurons extend dendrites towards the molecular layer and project axons through the hilus. By contrast, dendrites from seizure-induced neurons grow ectopically towards the hippocampal hilus instead of the molecular layer (Jessberger et al., 2007; Hattiangady and Shetty, 2010). In addition, a significant decline in neurogenesis is observed in the aging brain (Jessberger and Gage, 2008; Villeda et al., 2011; Spalding et al., 2013). These changes have been linked to depletion in the neural progenitor pool, owing to reduced Notch signaling, and an increase in astrogliogenesis at the expense of producing neurons (Lugert et al., 2010; Encinas et al., 2011). Finally, hippocampal neurogenesis deficits have been linked to cognitive defects characteristic of depression (Patricio et al., 2013), impairment in early Alzheimer's disease pathology (Demars et al., 2010; Faure et al., 2011; Mu and Gage, 2011; Rodriguez et al., 2011), bipolar disorder (Valvezan and Klein, 2012; Walton et al., 2012) and schizophrenia (SCZD) (Tamminga et al., 2010; Walton et al., 2012; Hagihara et al., 2013). More specifically, alterations in DG granule neuron maturation have been implicated in the etiology and pathogenesis of schizophrenia and bipolar disorder (see Box 1; Reif et al., 2006; Yamasaki et al., 2008; Walton et al., 2012; Hagihara et al., 2013; Shin et al., 2013).

To date, it has been difficult to investigate the role of hippocampal neurogenesis in the early stages of these CNS disorders in the human system, and it is not clear whether findings from studies using rodent models will translate across species. Recent advances in reprogramming technology now offer a renewable source of human neurons to investigate the basic biology of mechanisms of pathogenesis using cells that harbor the same genetic background and propensity for the disease as the patient. This technology provides the opportunity for drug screening using disease-relevant neurons in the hope that new therapeutics to target neurological pathologies such as SCZD may be developed. In addition, reprogramming technology can be used to develop in vivo cell therapies (for a review, see Yu et al., 2013). However, much additional work is needed to make cell transplantation a useful therapy. Although a number of preclinical studies using transplantation of stem cell-derived neurons in animals models have demonstrated functional improvements (Lu et al., 2003; Roy et al., 2006; Wernig et al., 2008; Hargus et al., 2010; Jiang et al., 2011; Ma et al., 2012), elucidating the precise mechanisms of functional recovery is still crucial for designing efficient and safe human trials.

In this Primer, we will discuss the advances that have been made in the past two decades to elucidate the process of DG development and identify the key regulators that drive DG granule neuron specification. In addition, we will discuss how these developmental principles have guided efforts to generate neural progenitor cells (NPCs) from pluripotent and somatic cells in vitro and, from these, mature DG granule neurons. Finally, we will discuss the current limitations and challenges of these approaches, as well as their suitability for disease modeling and therapeutic applications.

The formation of the central nervous system (CNS) requires precise orchestration of proliferation, cell fate determination and differentiation. Following gastrulation, cells derived from the mesoderm form the notochord and define the primitive axis of the embryo. The notochord then induces the formation of the neural plate, which later develops into the neural tube during primary neurulation (Colas and Schoenwolf, 2001). Following formation of the neural tube, neuroectodermal cells along the neural tube are patterned by gradients of signaling molecules along the dorsoventral and anterior-posterior axes to define regions of the forebrain, midbrain, hindbrain and spinal cord. The dorsal region of the forebrain, called the telencephalon, gives rise to the cerebral cortex, which comprises the neocortex, plaeocortex and archicortex, also known as the hippocampus, whereas the ventral telencephalon forms the basal ganglia. Dorsoventral and rostrocaudal differences are established by diffusible morphogens released from distinct locations, the temporal and spatial dynamics of which are crucial in patterning the distinct regions of the brain. Well-studied examples of these morphogens include fibroblast growth factor 8 (FGF8), which is produced rostrally at the anterior neural ridge, wingless (WNT3A, WNT2B, WNT5A) and bone morphogenetic proteins (BMPs), which are released dorsomedially by the cortical hem, and sonic hedgehog (SHH), which is released ventrally by the floor plate (Shimogori et al., 2004; Hebert and Fishell, 2008) (Fig. 1A). For the purpose of developing a differentiation protocol to generate the DG granule neurons, we focus our discussion on the role of these factors in the generation and patterning of the hippocampal DG and cornu ammonis (CA) subfields.

The hippocampus arises from the caudomedial edge of the dorsal telencephalic neuroepithelium adjacent to the cortical hem, a transient structure found during development after the invagination of the telencephalic roof at the dorsal midline (Lee et al., 2000). The hippocampal primordium is initially populated by Cajal-Retzius cells and radial glial cells, which proliferate in the dentate neuroepithelium (DNe). As the radial glial cells proliferate to expand the tissue, they produce embryonic neural stem cells (NSCs) that will later migrate to the SGZ of the DG to give rise to the prospero homeobox 1 (PROX1)⁺ DG granule neurons (Altman and Bayer, 1990a,b) (Fig. 2). The DG has a protracted developmental period that extends from embryonic day 13 (E13) to postnatal day 15 (P15) in mice, and begins in the dorsomedial region of the telencephalic neuroepithelium that directly abuts the cortical hem (Machon et al., 2007). The close proximity to the cortical hem is the first indicator that WNT activity features prominently in DG development and that the medial-lateral gradient of WNT signaling is essential for determining the cellular specificity in the hippocampal DG and CA subfields. Indeed, a number of studies have already demonstrated that WNT/β-catenin function is crucial for the correct development of the hippocampus. Mice deficient in Wnt3a, which encodes canonical WNT signaling protein expressed by the cortical hem, displayed severe defects in the size and organization of hippocampal DG (Lee et al., 2000). A mutation in Lef1 (lymphoid enhancer binding factor 1), the nuclear mediator of the WNT pathway, resulted in impaired granule cell development (Galceran et al., 2000), whereas a deficiency in the WNT co-receptor LRP6 (low density lipoprotein receptor-related protein 6) produced a marked reduction in the number of Prox1⁺ granule neurons in both the migratory route to the dentate and within the DG itself (Zhou et al., 2004). In addition to these loss-of-function phenotypes, gain-of-function approaches showed that sustained canonical WNT activity in the lateral and medial cortical primordium in D6-CLEF mutant mice resulted in ectopic generation of Prox1⁺ cells in the dorsomedial cortex at the expense of cortical neurons, providing further evidence that induction of the DG granule cell fate requires high levels of canonical WNT signaling (Machon et al., 2007).

In addition to WNT activity, SHH signaling has been shown to be involved in DG development, where it is important in expanding the granule neural progenitor population during perinatal development (Machold et al., 2003). At early postnatal periods, expression of Shh and its receptor patched (Ptch1) can be found in CA3, the hilar region of the DG and along the fimbria fibers that project to the hippocampus (Lai et al., 2003; Machold et al., 2003). Embryonic ablation of smoothened (Smo), an essential component of the SHH signaling, resulted in abnormalities in regions of postnatal neurogenesis, including a reduction in the number of granule cells in DG relative to other regions of the hippocampus (Machold et al., 2003; Han et al., 2008). Pharmacological inhibition of SHH signaling with cyclopamine also reduced the cell proliferation in the DG (Lai et al., 2003). Conversely, elevated hedgehog signaling produced increased cell proliferation in the DG, further confirming the mitogenic effects of SHH in vivo (Lai et al., 2003; Machold et al., 2003; Han et al., 2008). Interestingly, it appears that the SHH signaling required for DG proliferation is mediated via primary cilia. Mice lacking cilia genes, for example Kif3a (kinesin family member 3A) and Ift88 (intraflagellar transport 88), showed defects in adult neural progenitors, but rescue of DG proliferation with a constitutively activated Smo failed to restore neurogenesis in these mutants (Han et al., 2008).

Although it is clear that certain morphogens such as SHH and WNT are important for DG development, it is essential to consider the function of these pathways at different developmental time points. For example, for both WNT and SHH signaling, mutations in these pathways during an earlier developmental window produce much more substantial alterations in the regionalization of the brain than if the perturbations were introduced at a later time. Disruption of WNT or SHH signaling at E8-E10 in mice has been shown to result in strong ventralization or dorsalization of the telencephalon, respectively, whereas changes after E12 brought about alterations mostly attributed to a failure of neural progenitors to proliferate adequately (Chiang et al., 1996; Rallu et al., 2002; Backman et al., 2005). In addition, the interactions between the various signaling pathways active within the neurogenic niche must be closely examined. It has been shown that both SHH and WNT3A regulate the NSC population in the perinatal and adult DG through Sox2-dependent autocrine mechanisms (Favaro et al., 2009) and that each may mediate the proliferation and differentiation in different populations of neural progenitors in the DG during early postnatal and adult neurogenesis (Choe and Pleasure, 2012). These nuances in the function of developmental cues need to be carefully considered when taking on the challenge of recapitulating them in the in vitro setting.

Work is ongoing to elucidate the cellular and molecular mechanisms that regulate the different stages of neurogenesis during embryonic hippocampal development (Li and Pleasure, 2007). A number of proneural genes encoding transcription factors of the basic helix loop helix (bHLH) class that were previously identified for their role in neocortical development have recently been shown to also mediate cell fate specification and differentiation during DG neurogenesis (Fig. 3). Pro-neurogneic bHLH genes are activated downstream of signaling cues and have traditionally been divided into two groups: neuronal determination genes that are involved in the initial commitment to the neuronal fate and neuronal differentiation genes that regulate later events during differentiation to confer the identity of the terminally differentiated neuron (Pleasure et al., 2000).

The proneural transcription factor Ascl1 (achaete-scute complex homolog 1; Mash1) was one of the first neuronal determination genes to be identified and is expressed in hippocampal progenitors during development at different locations along the hippocampal primordium (Pleasure et al., 2000). However, although Ascl1 has been reported to promote acquisition of neuronal fate by inhibiting astrogliogenesis, Ascl1-null mice do not show significant defects in the proliferation of neural progenitors or generation of granule neurons in the DG (Nieto et al., 2001). Interestingly, upregulation of Ascl1 expression in adult DG progenitors promotes oligodendrocyte fate acquisition (Jessberger et al., 2008). Similar to Ascl1, neurogenin 2 (Ngn2) has been identified as a pro-neurogenic gene in cortical development, and its expression in the developing hippocampus overlaps with that of Ascl1 to a great extent (Galichet et al., 2008). However, Ngn2 mutant mice have reduced numbers of DG neural progenitors in the DNe, the dentate migratory stream and the tertiary matrix, resulting in the complete loss of the lower DG blade at birth. In addition, analysis of the radial glial cells in the DG of Ngn2 mutant mice showed an absence of GFAP⁺ radial glial fibers across the developing DG, likely contributing to the abnormal distribution of cells and DG morphology at birth (Galichet et al., 2008).

In addition to regulating the proliferation of DG neural progenitors, bHLH transcription factors can modulate the dentate migratory stream. Emx2 (empty spiracles homeobox 2) is a neuroepithelial transcription factor that was previously reported to play an important role in the development of the telencephalon by delineating the regionalization of the choroid plexus (von Frowein et al., 2006). However, it has been observed that Emx2 mutant mice also lacked the entire DG granule cell layer at birth, although the corresponding neuroepithelium appeared to be correctly specified (Savaskan et al., 2002). Follow-up studies showed that, in addition to significant decreases in cell proliferation in the DNe, the dentate migratory stream was almost entirely absent in the Emx2 mutants, resulting in the accumulation of Notch1⁺ dentate migratory precursor cells in the germinal layers around E15, and the ectopic presence of Neurod1⁺ and Prox1⁺ cells in the ventricular and subventricular zones at E18.5 (Oldekamp et al., 2004).

The T-box transcription factor Tbr2 (Eomes) has been found to regulate glutamatergic neuron fate commitment in several brain regions, including the DG (Hodge et al., 2012b). Loss of Tbr2 in the intermediate neuronal progenitor (INP) population during DG neurogenesis resulted in decreased granule cell neurogenesis due to INP depletion (Hodge et al., 2012b). This phenotype appears to be similar to that observed in the embryonic neocortex, where the loss of Tbr2 in INPs was shown to affect INP proliferation, leading to reduced production of neocortical pyramidal neurons (Arnold et al., 2008; Sessa et al., 2008). In addition, Tbr2 has been shown to promote neuronal lineage progression from NSC to INP, in part via the repression of Sox2 [SRY (sex determining region Y)-box 2], a key determinant of NSC identity (Hodge et al., 2012b). Interestingly, Tbr2 was recently found to also be expressed in Cajal-Retzius cells in the developing DG, and it is important for establishing and maintaining the radial processes of radial glial cells to facilitate the migration of newborn neurons (Hodge et al., 2013).

The specification of DG granule neuronal fate is distinct from other classes of DG or glutamatergic neurons and depends on activity of Prox1, a homeobox transcription factor homolog of the Drosophila melanogaster gene prospero. While Prox1 expression can be transiently found in several regions of the embryonic brain during development, its expression in the forebrain is highly restricted to the granule cell layer of the DG, starting at late gestation and persisting through adulthood (Torii et al., 1999; Lavado and Oliver, 2007). Furthermore, even though there are overlaps in the mechanisms that regulate cell fate and neuronal differentiation between the neocortex and hippocampal DG during development, the co-expression of Prox1 and Neurod1, a bHLH transcription factor that is required for neuroblast differentiation, is uniquely localized to the DG in late gestation at E18.5 (Liu et al., 2000; Lavado and Oliver, 2007; Gao et al., 2009). More than just a marker, Prox1 expression has been shown to be crucial for DG granule cell fate and can even instruct DG granule neuron identity at the expense of CA3 pyramidal neurons. In a recent study, overexpression of Prox1 in immature CA3 neurons resulted in the acquisition of DG granule neuron properties, demonstrating the important role of Prox1 as a postmitotic cell fate determinant for DG granule cells over CA3 pyramidal cells (Iwano et al., 2012). Indeed, Prox1 is essential for the maintenance of granule cell identity throughout life, as elimination of Prox1 in postmitotic immature DG neurons results in manifestation of CA3 pyramidal neuron properties, and the continual presence of Prox1 is required in mature DG cells for correct gene expression (Lavado et al., 2010; Iwano et al., 2012).

Similar to the differential roles observed for morphogens at different developmental stages, transcription factors in the hippocampus may exert temporally distinct roles in the regulation of DG neurogenesis depending on the developmental time point (Figs 1 and 3). For example, prior to E10.5, the LIM-homeodomain transcription factor LHX2 is important for defining the boundary of the cortical hem. Conditional knockout of Lhx2 results in an expanded cortical hem region at the expense of the neocortical epithelium, as well as mis-expression of WNT3A that leads to the formation of ectopic hippocampi (Mangale et al., 2008). However, at later stages, LHX2 regulates the cell fate decisions of hippocampal progenitors. Downregulation of Lhx2 in the embryonic hippocampus was found to promote astrogliogenesis at the expense of neurogenesis, and overexpression of Lhx suppressed GFAP promoter activity, thereby favoring neuronal differentiation (Mangale et al., 2008). The temporal regulation of neurogenesis and DG granule neuron development in particular must be considered to properly assess the correct specification of this cell type in vitro. Many studies have revealed different transcription factors that are involved in the generation and maturation of the granule DG neuron [i.e. FOXG1 (forkhead box G1), NEUROD1, TBR2, SOX11; for a review, see Hsieh, 2012). In addition, further characterization of temporal transcription factor expression dynamics during the directed differentiation process may provide greater insights into the underlying mechanisms involved in lineage specification and progression during in vivo DG granule neuron development.

The most intuitive approach to specifying the DG granule cell identity in vitro is to follow a stepwise lineage progression through an NPC stage, as occurs during in vivo development. Not surprisingly, the patterning cues that instruct NPC formation in vitro significantly overlap with those that have been identified in vivo. There have been multiple attempts to generate NPCs and neurons in vitro, mostly via directed differentiation of pluripotent stem cells (PSCs) but also more recently via direct reprogramming of somatic cells to NPCs and neurons via the forced activation of specific transcriptional networks (Kim et al., 2011; Lujan et al., 2012; Zhou and Tripathi, 2012; Wapinski et al., 2013; Ruggieri et al., 2014). These induced NPC and neuron (iNPC and iN) strategies can theoretically produce a more homogenous cell population, and may provide an advantage for translational studies by reducing the tumorigenic risk associated with transplanting pluripotent stem cell-derived cells in vivo. However, a current challenge for iNPC and iN approaches is the specification of neuronal subtypes. With the exception of one study that showed direct conversion of mouse and human fibroblasts into spinal motor neurons (Son et al., 2011), most direct reprogramming strategies produce functional iNs that appear to belong to multiple neuronal subtypes. In a single culture, these may include dopaminergic neurons marked by tyrosine hydroxylase (TH), glutamatergic neurons marked by vesicular glutamate transporter, interneurons marked by γ-aminobutyric acid, and motor neurons marked by choline acetyltransferase (Ambasudhan et al., 2011; Pang et al., 2011; Qiang et al., 2011; Yoo et al., 2011). In a similar way, neurons differentiated from iNPCs thus far also seem to contain multiple neuronal subtypes (Han et al., 2012; Ring et al., 2012; Wang et al., 2013; Zou et al., 2014). It is clear from these studies that the method of NPC production and the precise patterning cues that are provided have a direct bearing on the potential of the cells to reliably differentiate into specialized cell types. For this reason it is essential to evaluate the suitability of each of these methods not only in generating NPCs but also for correctly specifying the NPC population.

The unlimited self-renewal and largely unrestricted fate of PSCs makes them ideal candidates for generating specialized cell types in vitro. With the advent of iPSCs (Yamanaka and Takahashi, 2006; Yamanaka, 2007), there is also the possibility of creating disease-specific PSCs, opening up unparalleled opportunities for disease modeling and drug screening. Currently, there are only a few basic strategies for the directed differentiation of NPCs from PSCs, all of which rely on subtle variations in early patterning cues that occur during neuralization in vivo. Depending on exactly which patterning cues are provided in vitro, the resulting self-renewing NPC population can generate neurons enriched for a particular subtype, perhaps reflecting to some extent the endogenous NPC heterogeneity found in vivo (Perrier et al., 2004; Watanabe et al., 2005; Roy et al., 2006; Di Giorgio et al., 2008; Dimos et al., 2008; Eiraku et al., 2008; Gaspard et al., 2008; Marchetto et al., 2008; Kriks et al., 2011; Ma et al., 2011; Boyer et al., 2012; Shi et al., 2012; Vanderhaeghen, 2012; Maroof et al., 2013; Nicholas et al., 2013).

To specify a DG granule neuron subtype from NPCs, it is therefore essential to pattern the cells as per their normal in vivo development. Recently, Yu and colleagues (2014) did just this, using an embryoid body approach to recapitulate key morphological cues that specify hippocampal DG identity during development. Previously, the formation of free-floating EBs under serum-free conditions had been demonstrated to be particularly useful for the generation of naive telencephalic neural progenitors for the purpose of obtaining cortical neurons (Watanabe et al., 2005; Eiraku et al., 2008; Gaspard et al., 2008; Shi et al., 2012). These studies demonstrated that three-dimensional aggregates of the ESCs could self-organize to form pallial tissue under serum-free conditions (Eiraku et al., 2008; Kadoshima et al., 2013). Furthermore, this culture system could be patterned to enrich for telencephalon precursors by treatment with antagonists of WNT and NODAL pathways (Watanabe et al., 2005), and directed to impose dorsal attributes via antagonism of SHH signaling (Gaspard et al., 2008) during the early stages of EB formation. Building on these studies, Yu and colleagues developed a protocol for DG granule neuron differentiation that generated hippocampal NPCs, either entirely via EB formation or via EB formation followed by NPC monolayer formation. To generate the NPCs, the authors first mimicked the anterior-posterior patterning of the forebrain with factors such as DKK1 (dickkopf homolog 1), NOGGIN and the TGFβ inhibitor SB431542, to instruct the formation of telencephalic neural precursors. Antagonism of SHH signaling further enriched for dorsal forebrain progenitors. Overall, the treatment regime promoted upregulation of the pro-neurogenic genes found during early stages of DG neurogenesis, such as Sox2, Pax6 (paired box 6), Emx2 (empty spiracles homeobox 2) and Foxg1, which were expressed in a pattern that mimicked in vivo DG development (Fig. 4) (Yu et al., 2014). As will be discussed in the section below on DG granule neuron specification, these cells gave rise to Prox1⁺ DG granule neural cells, demonstrating the importance of recapitulating key developmental pathways to correctly pattern PSC-derived NPCs.

To date, there has been only one study demonstrating the in vitro generation of DG granule neurons (Yu et al., 2014). Following the initial period of patterning designed to obtain the hippocampal NPCs, the cells were further treated with two key hippocampal neurogenic molecules: WNT3A and BDNF. WNT3A had previously been shown to be important for the maintenance of hippocampal NPCs and their differentiation into DG granule neurons, whereas BDNF is a neurotrophic factor that had been previously shown to promote hippocampal neurogenesis throughout life (Fig. 4) (Scharfman et al., 2005; Erickson et al., 2010). Quantitative real-time PCR (qPCR) analysis of whole populations of immature neurons patterned in this way showed concomitant expression of Neurod1 together with Prox1, as is uniquely found in the developing DG in vivo (Yu et al., 2014), although whether Neurod1 and Prox1 were co-expressed in immature neurons at the single-cell level was not reported. Importantly, however, the mature neurons differentiated from these hippocampal NPCs triple-labeled for PROX1, a marker specific to DG granule neurons in the hippocampus (Hodge et al., 2012a; Iwano et al., 2012), as well as MAP2AB (microtubule-associated protein 2) and NEUN (RBFOX3; RNA binding protein, fox-1 homolog 3), markers of postmitotic neurons (Fig. 5A) (Hodge and Hevner, 2011). These human DG granule neurons could be dissociated from the EBs and maintained on co-cultures with hippocampal astrocytes for up to 3 months in vitro. Further electrophysiology and calcium imaging analyses demonstrated the maturation, connectivity and network properties of these cells in the astrocyte co-culture system (Fig. 5B). The isolated NPC population was also successfully transplanted into the developing DG of P10 animals to produce electrophysiologically active Prox1⁺ neurons. These neurons were able to functionally integrate into the endogenous neural circuitry and produce glutamine-mediated postsynaptic currents following stimulation of the perforant path, the main afferent input into the DG from the entorhinal cortex (Fig. 5C) (Witter, 2007). The fact that the PSC-derived NPCs were able to functionally integrate and produce DG granule neurons in vivo supports the hypothesis that the NPCs were correctly patterned in vitro and highlights the importance of recapitulating regionally specific patterning of NPCs to produce specific neuronal subtypes in vitro.

Stem cell biology has opened up new avenues in the field of neuroscience research as well as for therapeutic approaches for neurological diseases. Human ESCs are fundamental tools for the study of human neurodevelopment in an in vitro setting, providing glimpses into crucial developmental windows via their differentiation process. In addition, cellular reprogramming has made it possible to modify somatic cells of adults and neonates into human iPSCs, NPCs and even neurons, thus providing a renewable source of human neurons that can be used to investigate the mechanism of pathogenesis and assess the safety and efficacy of candidate drugs.

Despite the promise of recent advances such as direct reprogramming and directed differentiation, considerable hurdles remain before cells produced via these strategies can be used for therapeutic purposes. Chief among these is a deeper understanding of the precise molecular and functional attributes of the in vitro-derived cells. In the example of generating the DG granule neurons, an ideal approach would require single-cell expression profile characterization from in vitro-derived cells followed by comparison with endogenous newborn and adult granule neuron cell profiles from living animals. Further immunocytochemistry at select timepoints may provide additional evidence for the developmental progression of PSC-derived DG granule neurons. Such comparisons could help to identify specific populations that correspond to distinct DG granule neuron developmental timepoints in vivo, and enable selection of these cell types in vitro. Another tool for characterizing cellular subtypes, translating ribosome affinity purification (TRAP) has been reported to be useful for obtaining the translated mRNA profiles in genetically defined cell populations (Doyle et al., 2008; Heiman et al., 2008), and has recently facilitated the identification cortical neuron subtypes that mediate responses to chronic antidepressant treatments (Schmidt et al., 2012). Indeed, characterization of neuronal subtypes, both of endogenous origin and their counterparts produced in vitro, is an ongoing effort that will continue to shape the stem cell field by redefining the concept of cellular plasticity and the properties, both cell-autonomous and non-cell-autonomous, that confer cellular subtype specificity.

As cognitive dysfunction associated with aberrant hippocampal neurogenesis has been implicated in CNS disorders, generation of DG granule neurons from patient cells may offer a unique opportunity for disease modeling and drug discovery (Jun et al., 2012). Although translation of these findings to a clinical setting is still far away, it is still possible to gain insight into the physiological processes that manifest in certain neuropathological diseases, perhaps hinting at possible targets for pharmaceutical intervention. The recent study by Yu and colleagues reported reduced neuronal network activity by calcium imaging and electrophysiological recording in DG granule neurons generated from reprogrammed schizophrenia patient-derived fibroblasts, compared with the same neuronal sub-type generated from normal human fibroblasts. Importantly, the disease phenotype was not observed under a pan-neuronal differentiation protocol, but only manifested when the correct neuronal sub-type was specified (Brennand et al., 2011; Yu et al., 2014) (see Box 1 on role of DG granule neurons in schizophrenia). This finding supports the concept that selective enrichment of clinically relevant neuronal subtypes is an important strategy to improve the accuracy of disease modelling. As our understanding of neuronal sub-type specification continues to evolve with new insights gained from developmental studies, so too must the differentiation strategy used to generate the desired cell type. For example, recent work from Choe and colleagues showed that BMP7, a member of the bone morphogenetic proteins family, is crucial for initiating neurogenesis in the DG progenitors by signaling through the activin receptor type I (ACVR1) and inducing Lef1 expression (Choe et al., 2013). As both BMP7 and WNT3A are present in high concentrations and mediate the Lef1 expression during DG development, it would be very interesting to examine the potentially synergistic effects of these two molecules on the production of Tbr2⁺ intermediate neural progenitors and the subsequent Prox1⁺ neurons during DG granule neuron differentiation in vitro.

In this Primer, we have discussed the development and design of approaches to generate hippocampal DG granule neurons in vitro for the purposes of disease modeling and drug screening. The differentiation process recapitulates important aspects of DG neurogenesis and provides the opportunity to examine the etiology and pathogenesis of diseases in developmental time. However, this is only the first step in the process of developing an even more complete platform to model the complexity of the dynamic connections that exist between different classes of neurons in the hippocampus niche and to investigate the cellular and molecular mechanisms underlying hippocampus-related disorders. The fundamental principle of recapitulating DG developmental cues to generate DG granule neurons could also be used to develop additional differentiation approaches to produce the various subtypes of pyramidal neurons found in the CA subfields. If multiple different neuronal subtypes can be successfully generated in vitro, then this may provide the opportunity to recapitulate neural circuits via engineered platforms, e.g. by using microfluidics devices with microfabricated substrates (see Box 2). This approach would facilitate the dissection of the function and interactions between subpopulations of neurons under physiological and pathophysiological conditions (for additional reading, see Yu et al., 2013). Such high-resolution in vitro modeling of neural circuitry will allow unprecedented control and insight into how the mammalian brain functions and how this function is derailed during the progression of neuropathological disease.

The authors thank M. L. Gage for editorial comments.