Functional properties of in vitro excitatory cortical neurons derived from human pluripotent stem cells

Abstract The in vitro derivation of regionally defined human neuron types from patient‐derived stem cells is now established as a resource to investigate human development and disease. Characterization of such neurons initially focused on the expression of developmentally regulated transcription factors and neural markers, in conjunction with the development of protocols to direct and chart the fate of differentiated neurons. However, crucial to the understanding and exploitation of this technology is to determine the degree to which neurons recapitulate the key functional features exhibited by their native counterparts, essential for determining their usefulness in modelling human physiology and disease in vitro. Here, we review the emerging data concerning functional properties of human pluripotent stem cell‐derived excitatory cortical neurons, in the context of both maturation and regional specificity.


Introduction
The human cerebral cortex controls higher cognitive abilities, including those that distinguish human beings from other mammalian species such as abstract thinking and complex language (Defelipe, 2011). This highly complex structure consists of several distinct areas that are responsible for processing and integrating different motor and sensory information and information stored in memory. Further functional and anatomical sophistication is reflected in the diversity and number of excitatory and inhibitory cell types, each characterized by distinct gene expression, morphology and functional properties (Molyneaux et al. 2007;Lui et al. 2011). The cortex is a target for many disorders of the brain at all stages of life. For example, perturbation of cortical development can lead to neurodevelopmental disorders, such as autism spectrum disorders, while the adult cortex is a major site for certain neurodegenerative conditions including Alzheimer's disease, frontotemporal dementia and Huntington's disease.
The majority of studies investigating cortical development and neuronal function have been based upon rodent models, but several aspects of the human cortex differ greatly from rodents including tangential expansion in the frontal cortex and gyrencephaly (Defelipe, 2011;Lui et al. 2011). Recently many laboratories have established in vitro protocols to derive excitatory pyramidal neurons, the principal neuronal type in the adult cortex, from human pluripotent stem cells (hPSCs) that provide powerful, readily accessible tools to model neuronal function in both healthy and disease contexts van den Ameele et al. 2014).
The ability of neurons derived from mouse and more recently from human stem cell lines to exhibit classical neuronal functional properties both in vitro and when integrated into host systems has been studied for many years (Benninger et al. 2003;Wernig et al. 2004). Increasing refinement and control in the derivation of neurons generated from embryonic or induced pluripotent stem cells (ESCs/iPSCs) now means it is possible to derive regionally specific neurons, including hPSC-derived excitatory cortical neurons (hereafter termed 'hPSC C neurons'). This raises the question as to the extent such cells are able to recapitulate known details of native cortical development and ultimately whether they are appropriate models of 'diseases in a dish' (Sandoe & Eggan, 2013). In addition to forming synaptic connections to generate the intricate circuitry responsible for complex cortical processes, native excitatory cortical neurons undergo distinctive developmental changes in ion channel expression and ionic gradients that determine their function within cortical networks (Moody & Bosma, 2005). This review will principally focus on the emerging data examining the functional capability of hPSC C neurons to exhibit known native-like properties with regard to functional maturation and regional specification.

Summary of cortical development
The cerebral cortex organizes in a complex 3-dimensional structure comprising several anatomically distinct cortical areas. Cortical regionalization within the developing telencephalon is determined by graded expression of transcription factors expressed by cortical progenitor cells, including Gli3, Emx2, Pax6, Sp8 and Coup-TF1 (O'Leary et al. 2007). This, in turn, is established by three major telencephalic signalling centres: the cortical hem that secretes Wnt and bone morphogenetic protein, the anterior neural ridge that secretes Fgf8, and the ventral telencephalon that secretes sonic hedgehog (Shh). Once the major cortical axes have been established, cortical progenitor cells generate neurons through the process of cortical lamination in a time-dependent manner (Gaspard & Vanderhaeghen, 2011). During this process, later-born neurons migrate over the early-born neurons within the cortical plate forming a layered structure in an inside-out fashion. Therefore, the deepest cortical layer VI forms first and the upper layer II last; the marginal zone or layer I, containing reelin + Cajal-Retzius cells, escapes this inside-out process (Frotscher, 1998). The adult cortex has a structure made up of six defined layers (I-VI) of diverse excitatory cortical neuron types that can be identified by layer-specific markers, including CUX1 (layer II/III), ROR-β (layer IV), CTIP2 (layer V), SOX5 (layers V and VI) and TBR1 (layer VI) (Molyneaux et al. 2007).
The correct time-dependent patterning of the telencephalon and the formation of different neuronal subtypes within the different layers are ultimately essential for the appropriate functional wiring of cortical neurons with other cortical or subcortical targets (Germain et al. 2013). While neocortical excitatory pyramidal neurons form from distinct populations of cortical progenitor cells within the embryonic dorsal telencephalon, the majority of inhibitory interneurons originate in the ventral telencephalon and subsequently tangentially migrate into the developing cortex (Hansen et al. 2013).

Specification of in vitro hPSC C cultures
In vitro-derived cortical progenitor cells giving rise to hPSC C neurons can be formed from the spontaneous neuralization of hPSCs (Eiraku et al. 2008;Li et al. 2009;Mariani et al. 2012;Shi et al. 2012) and neuralization can be accelerated by using inhibitors of Wnt and bone morphogenetic protein/NODAL signalling pathways (dual-SMAD inhibition; Chambers et al. 2009). Moreover, the presence of Shh or Shh agonists can give rise to ventral telencephalic progenitors from which cortical interneurons can differentiate (Germain et al. 2013;Maroof et al. 2013). Furthermore, hPSC-derived cortical progenitors generally acquire a caudal identity by default as shown by their pattern of projections when transplanted in mouse brains (Espuny-Camacho et al. 2013), yet they can be patterned to different cortical regions and respond to signalling cues when treated with morphogen agonists (Eiraku et al. 2008;Espuny-Camacho et al. 2013;Kadoshima et al. 2013). Moreover, whilst the temporal generation of neurons belonging to different layers is largely maintained in vitro and the presence of neurons belonging to all six layers has been reported, the contribution of each layer considerably varies depending on the method used (see van den Ameele et al. 2014).
In these respects, in vitro neuronal connections and circuitry will therefore be somewhat limited by the protocol to recapitulate cortical development in its cellular specification and organization, particularly in monolayer cultures. However, aspects of cortical cytoarchitecture are remarkably maintained in vitro, in particular in 3-dimentional cultures that allow the radial localization of later-born neurons above earlier-born ones (Mariani et al. 2012;Lancaster et al. 2013;Paşca et al. 2015).
Transcriptome analyses of the sequential phases of in vitro hPSC-derived corticogenesis in monolayer or 3-D culture demonstrate the ability of these protocols to reproduce gene expression profiles in hPSC C neuron populations that are equivalent to early native human embryonic cortical development (Stein et al. 2014;Paşca et al. 2015). Importantly, the heterogeneity in the cellular specification of previously published protocols is also matched at the transcriptome level, where some protocols better recapitulate native development than others and, furthermore, have variable rates of hPSC C neuron maturation (Stein et al. 2014). These data, however, do not directly examine the functional capacity of hPSC C neurons. Many groups therefore apply patch-clamp electrophysiology (and also live-imaging) to assess directly the functional membrane properties of hPSC C neurons.

hPSC C neurons are functionally reminiscent of immature cortical neurons
Neurons are defined by their excitable plasma membrane properties, which rely on the development of ionic gradients and the expression of ion channels. Reports demonstrating the generation of hPSC C neurons have therefore focused upon basic intrinsic membrane properties and the ability of hPSC C neurons to generate action potentials. Data indicate that intrinsic and action potential firing properties of hPSC C neurons are broadly comparable to rodent cortical neurons at an embryonic/early postnatal stage of development (Johnson et al. 2007;Kim et al. 2011;Espuny-Camacho et al. 2013;Bilican et al. 2014). Indeed, the majority of studies of hPSC-derived neurons which have recorded electrophysiological properties consistently report input resistances that are 5-to 20-fold higher than their adult, mature in vivo counterparts (Johnson et al. 2007;Kim et al. 2011;Espuny-Camacho et al. 2013;Bilican et al. 2014). This is also true for hPSC-derived neurons cultured for extended periods. Thus, hPSC-derived neurons typically require only modest current injection to elicit action potential firing. hPSC C neurons that show increases in excitability over time also demonstrate maturation in their expression of intrinsic membrane conductances that collectively define the nature of neuronal excitability (Johnson et al. 2007;Bilican et al. 2014). Additionally, interneuronal content is very low in some in vitro cultures (Shi et al. 2012;Bilican et al. 2014); however, many protocols do not report the extensiveness of interneuron differentiation in their cultures. It is important to note that interneuronal classes have highly variable mature firing properties (Markram et al. 2004) and as such this may be confused with firing patterns seen with immature excitatory neurons.
Studies have revealed that the resting membrane potential (RMP) of hPSC C neurons hyperpolarizes with extended culture periods (Johnson et al. 2007;Bilican et al. 2014). Nevertheless RMPs can remain relatively depolarized and as such this compromises their ability to display spontaneous (TTX-sensitive) action potential firing which is considered critical to the development and maturation of the cortex (Spitzer, 2006). Indeed, Weick et al. (2009) demonstrated that TTX only weakly blocked spontaneous Ca 2+ transients in hPSC C neuron cultures indicating that these transients were largely not J Physiol 594.22 mediated by spontaneous action potential activity. Careful pharmacological work in this study determined the source of Ca 2+ to be mediated through L-type voltage-gated Ca 2+ channels and transient receptor potential channels.

Ligand-gated ion channels
Ligand-gated ion channels (LGICs) are integral to the process of fast neurotransmitter signalling and their activities contribute to the fine balance of excitation and inhibition within the CNS.
LGICs are multimeric protein complexes that can comprise numerous subunit combinations which impose distinct biophysical and pharmacological properties.
LGIC subunit composition is often regulated both developmentally and spatially. Thus, the assessment of LGIC composition in hPSC C neurons is essential if we are to determine the extent to which these cells reflect native properties.

Ionotropic glutamate receptors
Ionotropic glutamate receptors are the central mediators of fast excitatory neurotransmission in the cortex and are a family of three tetrameric receptor types: AMPA, NMDA and kainate receptors (Traynelis et al. 2010). NMDA receptors (NMDARs) are composed of two ubiquitously expressed GluN1 subunits and two potential GluN2A, GluN2B, GluN2C, GluN2D and/or GluN3A, GluN3B subunits . Considerable evidence shows that NMDARs in embryonic mammalian cortical neurons contain predominantly GluN1 and GluN2B subunits while maturation is associated with a functional up-regulation of GluN2A subunits . For rodents this is a postnatal event; however, determining this in humans has proved challenging (Henson et al. 2008). hPSC C neurons maintained in culture for 5 weeks express GluN1/GluN2B NMDARs as assessed by their sensitivity to the GluN2B-selective antagonist ifenprodil and therefore an immature NMDAR profile .
AMPA receptors (AMPARs) can be composed of GluA1, GluA2, GluA3 and GluA4 subunits of which the functional up-regulation of the GluA2 subunit is associated with cortical neuronal maturation (Traynelis et al. 2010). Furthermore, GluA2 subunits predominantly undergo post-transcriptional modification resulting in an arginine codon (GluA2(R)) replacing a glutamine codon (GluA2(Q)) in the M2 pore-forming region of the channel. Therefore native cortical maturation is associated with a shift from GluA2(R)-lacking to GluA2(R)-containing AMPARs. Importantly, the presence of one or more GluA2(R) subunit in an AMPAR complex results in reduced single-channel conductance, reduced sensitivity to channel-blocking polyamines and, crucially, reduced Ca 2+ permeability (Traynelis et al. 2010). Assessment of the functional AMPAR composition in hPSC C neurons using non-stationary fluctuation analysis to estimate mean AMPAR single-channel conductance and their sensitivity to a GluA2(R)-lacking AMPAR channel blocker indicates an activity-independent and native-like maturation from GluA2(R)-lacking to GluA2(R)-containing AMPARs within 5 weeks of in vitro differentiation . GluA2 transcript expression also increases with time in culture (Chander & Weick, 2014;Stein et al. 2014;van de Leemput et al. 2014). Thus AMPAR expression in hPSC C neurons appears to display properties that are observed in native mature neuronal populations (Isaac et al. 2007). Both NMDARs and AMPARs are expected to undergo maturational changes in composition in the early postnatal weeks of cortical development in rodents (Traynelis et al. 2010) and in this regard the ontogenetic development of AMPARs in hPSC C neurons is much more rapid than expected. Interestingly, the GluA2 subunit has been shown to be rapidly edited and functionally up-regulated 4 weeks after the in vitro differentiation of neurons from primary human cortical progenitors (Whitney et al. 2008) in contrast to the expected longer in vivo developmental time scales (Talos et al. 2006).
These data suggest the rapid maturation of the AMPAR complex is a potential product of the in vitro environment. Nonetheless, this feature provides an opportunity to examine numerous scenarios in which abnormal regulation of the GluA2 subunit is hypothesized or prevalent in adult human disease (Wright & Vissel, 2012).

Ionotropic GABA and glycine receptors
GABA A receptors (GABA A Rs) and strychnine-sensitive glycine receptors (GlyRs) are pentameric LGICs that primarily mediate fast inhibitory neurotransmission in the mature cortex. GABA A Rs can be potentially composed of 19 known subunits (α1-6, β1-3, γ1-3, δ, ε, π, θ and ρ1-3) giving a vast number of theoretical possible GABA A R arrangements. Whilst in reality composition is tightly regulated, the breadth of possible GABA A R composition generates considerable functional and pharmacological diversity across brain regions and cellular locations (Olsen & Sieghart, 2009). Using a pharmacological and RNA-seq-based approach hPSC C neurons differentiated for 5 weeks were shown to express GABA A Rs that had a predominant α2/3β3γ2 composition (James et al. 2014). This is the most common GABA A R combination present in the embryonic cortex (Olsen & Sieghart, 2009). Comparison of GABA A R subunit transcript levels in hPSC C neurons with data from human primary tissue indicates that overall GABA A R subunit expression is similar to that seen in the cortex at 12-21 weeks postconception. Again this is in broad agreement with other transcriptome-based studies (Stein et al. 2014;Paşca et al. 2015). Finally, pharmacological assessment of GABA A Rs in hPSC C neurons is consistent with the absence of the α1-subunit which is associated with more developmentally mature cortical neurons.
GlyRs are thought to play an important role in cortical development and transient functional GlyR expression is a feature of neocortical development in rodents (Flint et al. 1998;Avila et al. 2013). Indeed, hPSC C neurons respond robustly to glycine application (James et al. 2014). Pharmacological and RNA-seq analysis of GlyRs indicates that the GlyR composition is principally α2/β-containing (James et al. 2014). Transcript levels indicate a level of maturity equal to that of GABA A Rs; however, it is thought that the early mammalian embryonic GlyR composition consists of homomeric α2 GlyRs and matures to α1/β-containing GlyRs (Lynch, 2009). Interestingly, a transient GlyR population of α2/β-containing GlyRs has been observed in developing rodent Cajal-Retzius cells (Okabe et al. 2004). hPSC C neurons may prove useful in elucidating the role of GlyRs within the developing human cortex.

Intracellular chloride
GABA A Rs and GlyRs are permeable to Cl − ions and mediate inhibitory responses in adult cortical neurons. However, the application of either GABA or glycine to embryonic excitatory cortical neurons generates depolarizing excitatory responses due to elevated levels of intracellular Cl − (Ben-Ari et al. 2007). The developmental reduction in intracellular Cl − concentration is a crucial feature of cortical development and its perturbation is implicated in numerous disease mechanisms (Blaesse et al. 2009). Intracellular Cl − activity in hPSC C neurons (and neural precursor cells) measured using the perforated patch-clamp technique falls from around 25 mM after 7 weeks in culture to <7 mM in hPSC C neurons . Correspondingly, the expression of K + -Cl − cotransporter-2 (KCC2, responsible for Cl − efflux) increases while that of the Na + -K + -Cl − cotransporter-1 (NKCC1, responsible for Cl − influx) falls in hPSC C neurons, which is in agreement with native developmental mechanisms that regulate intracellular Cl − activity (Ben-Ari et al. 2007). In agreement with our data, Shcheglovitov et al. (2013) also report that application of GABA to hPSC C neurons generates hyperpolarizing currents consistent with a reduction in intracellular chloride activity. The time of this switch has been reported with high variability in the human cortex (Blaesse et al. 2009), though it is interesting that a population of Pax-6 + neural precursors give rise to subplate neurons that highly express KCC2 by 16 weeks postconception (Bayatti et al. 2008) and that Livesey et al. (2014) studied hPSC C neurons from pre-dominantly Pax-6 + neural precursors. Livesey et al. (2014) also highlight that the use of neurotrophic media supplements influences the development of expression of Cl − transporters and Cl − activity. Although neurotrophic factors are added to promote synaptogenesis and/or increase responsiveness to neurotransmitters in pluripotent stem cell (PSC)-derived neuronal cultures (Copi et al. 2005;Bardy et al. 2015), neurotrophic factors have an important role in the regulation of Cl − transporter expression (Blaesse et al. 2009). Beyond pharmacological considerations regarding the use of chronic neurotrophic factor media supplements (Frank et al. 1996), the impact of neurotrophic factors on PSC-derived neuronal physiology needs to be carefully considered.

Synaptic and network properties
In addition to harbouring excitable membranes, it is a defining feature of neurons to receive and generate synaptic signals. These exist in two general forms; phasic and tonic, and are both essential to the normal function of the CNS. The intricate and specific synaptic connectivity displayed by native cortical neurons is key to cortical network development and function (Spitzer, 2006). It is therefore critical to the development of in vitro hPSC C -derived neurons to recapitulate native synaptic properties. The co-localization of pre-and postsynaptic membrane-associated scaffold proteins such as synaptophysin and PSD-95, respectively, provide an indication of architectural synapse formation, but not functionality. Functional synaptic activity is generally detectable in standard cultures of in vitro hPSC C -derived neurons where phasic ionotropic glutamatergic receptorand GABA A R-mediated activity has been observed. The latter property is determined by the culture protocol employed and its potential to generate GABA-ergic interneurons. However, it is largely accepted that many standard in vitro hPSC protocols do not generate cultures that exhibit robust synaptic activity (Bardy et al. 2015). Studies examining synaptic properties therefore employ techniques to promote synaptic formation in their cultures.
Moreover, many groups now co-culture hPSC C neurons with primary rodent astrocytes, which also promotes synaptic function (Johnson et al. 2007;Kim et al. 2011;Shcheglovitov et al. 2013;Wen et al. 2014;Pak et al. 2015). The analysis of the field-evoked postsynaptic events in co-cultured neurons demonstrates that glutamate activates both fast AMPAR-mediated and slower GluN1/GluN2B-like NMDAR-mediated events in control neurons (Shcheglovitov et al. 2013). Interestingly in this study, cell lines lacking the postsynaptic density protein SHANK3, which is reduced in expression in a neurodevelopmental disorder (Phelan-McDermid syndrome), exhibit evoked NMDAR-mediated postsynaptic currents that have faster decay kinetics, which may be consistent with an NMDAR population expressing both GluN2A and GluN2B NMDAR subunits. Alternatively, presynaptic neurotransmitter-release dysfunction has been modelled in hPSC C neurons derived from schizophrenia (mutant DISC1; Wen et al. 2014) and autism (mutant NRXN1; Pak et al. 2015) patients. Recently, Bardy et al. (2015) reported that relatively sparse synaptic activity observed in PSC-derived neurons is due, in part, to the media in which the neurons are maintained. Their study utilized a culture medium that promoted synapse formation and concomitantly increased functional AMPAR-mediated synaptic activity in PSC-derived neurons co-cultured with mouse astrocytes. Thus, hPSC C neurons can recapitulate synaptic activity and offer the exciting potential to study synaptic dysfunction to elucidate disease mechanisms. Table 1 describes reports of advanced culture techniques in order to study synaptic physiology. Similarly in in vitro cultures, reports of robust network activity have not yet been reported widely. A potential factor in the inability to observe synchronous multi-neuronal firing in many cultures may be a consequence of the extent to which GABA-ergic interneurons are present or absent in cultures. GABA-ergic signalling is established before that of glutamatergic-signalling in early development and is thought to initiate primitive neural network activity (Ben-Ari et al. 2007). Indeed, inhibitory GABA-ergic interneurons are essential for maintaining the balanced activity of cortical neural circuits. GABA-ergic synaptic activity has been detected within hPSC C cultures in the form of spontaneous postsynaptic currents (Johnson et al. 2007;Wu et al. 2007;Shcheglovitov et al. 2013). However, a recent report has observed synchronized excitatory neurotransmitter-driven network activity that resembles that of early-stage cortical development and which occurs in cultures that contain a low percentage of interneurons and is insensitive to pharmacological blockade of GABA A Rs (Kirwan et al. 2015). A key study in the near future will therefore be the culture of defined mixtures of hPSC C neurons and defined populations of GABA-ergic interneurons. Notably, it has been shown that hPSC C neurons can integrate into the network firing properties generated by primary mouse cells (Weick et al. 2011) and, furthermore, light-stimulated hPSC-derived neurons transduced with channelrhodopsin can influence the network activity of hippocampal organotypic slices (Piña-Crespo et al. 2012).

Future challenges: making mature neurons
The capacity of hPSC C neurons to become physiologically relevant is their successful incorporation into a native system. As an initial step along this road Espuny-Camucho et al. (2013) successfully demonstrated the differentiation and functional integration of hPSC C neurons into rodent cortex. Several months after transplantation into the developing cortex hPSC C neurons exhibited intrinsic membrane properties consistent with adult mature neurons, in contrast to the more immature properties of in vitro differentiated hPSC C neurons. hPSC C neurons incorporated into the rodent cortex also received synaptic inputs. Integration of hPSC C neurons expressing channelrhodopsin to the mouse hippocampal CA1 region has demonstrated the ability to generate light-evoked postsynaptic currents in adjacent (untransfected) neurons (Weick et al. 2011). These data indicate the potential of such neurons to contribute to cortical network activity. Furthermore, it has been reported that hPSC C neurons integrated into a rat model of stroke promote functional recovery (Tornero et al. 2013). While there is clearly still much to learn with regard to in vivo functional integration, studies such as these give strong support to the notion that hPSC C neurons have the capacity to possess functionally mature phenotypes. Similarly, hPSC C -derived ventral telencephalic interneurons develop mature intrinsic properties and receive synaptic input when integrated into the embryonic rodent cortex (Nicholas et al. 2013). Furthermore, this study indicates that interneuron maturation can be achieved with co-culture in the presence of rodent glia and requires extended culture periods (>6 months). An equivalent study has not been performed for hPSC C excitatory neurons. These data indicate that in vitro hPSC-derived neuron cultures lack important factors that are present in vivo and that are required for maturation.
Aside from their electrophysiological phenotype, hPSC C neurons need to display morphological characteristics typical of native cortical neurons. Dendritic spines form the principal location at which excitatory synaptic transmission and synaptic plasticity take place and, moreover, numerous neurological diseases are accompanied by spine number or size alterations (Bourne & Harris, 2008;Penzes et al. 2011). In this respect it is of significance that in vivo integrated hPSC C neurons do appear to develop spine-like structures (Espuny-Camucho et al. 2013), but dendritic spine structures with co-localized expression of PSD-95 are infrequent in in vitro hPSC-derived neurons (Marchetto et al. 2010).
In conclusion, a major challenge is to generate neuronal populations that exhibit maturation profiles that more closely reflect those seen in vivo. Co-culture with astrocytes and mixed neuronal populations, together with the maintenance of cells in media that promote increased synaptic activity, indicates that such strategies are required to assess synaptic transmission. When achieved, this will enhance and make more relevant our ability to study human physiology and pathophysiology using in vitro hPSC C neurons.